Type:	Package
Title:	Test Theory Analysis and Biclustering
Version:	1.5.1
Description:	Implements comprehensive test data engineering methods as described in Shojima (2022, ISBN:978-9811699856). Provides statistical techniques for engineering and processing test data: Classical Test Theory (CTT) with reliability coefficients for continuous ability assessment; Item Response Theory (IRT) including Rasch, 2PL, and 3PL models with item/test information functions; Latent Class Analysis (LCA) for nominal clustering; Latent Rank Analysis (LRA) for ordinal clustering with automatic determination of cluster numbers; Biclustering methods including infinite relational models for simultaneous clustering of examinees and items without predefined cluster numbers; and Bayesian Network Models (BNM) for visualizing inter-item dependencies. Features local dependence analysis through LRA and biclustering, parameter estimation, dimensionality assessment, and network structure visualization for educational, psychological, and social science research.
License:	MIT + file LICENSE
Language:	en-US
Encoding:	UTF-8
LazyData:	true
BuildVignettes:	true
URL:	https://kosugitti.github.io/exametrika/
RoxygenNote:	7.3.2
Suggests:	testthat (≥ 3.0.0)
Config/testthat/edition:	3
Depends:	R (≥ 4.1.0), mvtnorm, igraph
NeedsCompilation:	no
Packaged:	2025-03-09 04:59:57 UTC; napier3
Author:	Koji Kosugi [aut, cre]
Maintainer:	Koji Kosugi <kosugitti@gmail.com>
Repository:	CRAN
Date/Publication:	2025-03-09 05:10:02 UTC

Alpha Coefficient

Description

This function computes Tau-Equivalent Measurement, also known as Cronbach's alpha coefficient, for a given data set.

Usage

AlphaCoefficient(x, na = NULL, Z = NULL, w = NULL)

Arguments

Value

For a correlation/covariance matrix input, returns a single numeric value representing the alpha coefficient. For a data matrix input, returns a list with three components:

AlphaCov

Alpha coefficient calculated from covariance matrix

AlphaPhi

Alpha coefficient calculated from phi coefficient matrix

AlphaTetrachoric

Alpha coefficient calculated from tetrachoric correlation matrix

References

Cronbach, L. J. (1951). Coefficient alpha and the internal structure of a test. Psychometrika, 16,297–334.

Alpha Coefficient if Item removed

Description

This function returns the alpha coefficient when the specified item is excluded.

Usage

AlphaIfDel(x, delItem = NULL, na = NULL, Z = NULL, w = NULL)

Arguments

Bicluster Network Model

Description

Bicluster Network Model: BINET is a model that combines the Bayesian network model and Biclustering. BINET is very similar to LDB and LDR. The most significant difference is that in LDB, the nodes represent the fields, whereas in BINET, they represent the class. BINET explores the local dependency structure among latent classes at each latent field, where each field is a locus.

Usage

BINET(
  U,
  Z = NULL,
  w = NULL,
  na = NULL,
  conf = NULL,
  ncls = NULL,
  nfld = NULL,
  g_list = NULL,
  adj_list = NULL,
  adj_file = NULL,
  verbose = FALSE
)

Arguments

Value

nobs

Sample size. The number of rows in the dataset.

testlength

Length of the test. The number of items included in the test.

Nclass

Optimal number of classes.

Nfield

Optimal number of fields.

crr

Correct Response Rate

ItemLabel

Label of Items

FieldLabel

Label of Fields

all_adj

Integrated Adjacency matrix used to plot graph.

all_g

Integrated graph object used to plot graph.see also plot.exametrika

adj_list

List of Adjacency matrix used in the model

params

A list of the estimated conditional probabilities. It indicates which path was obtained from which parent node(class) to which child node(class), held by parent, child, and field. The item Items contained in the field is in fld. Named chap includes the conditional correct response answer rate of the child node, while pap contains the pass rate of the parent node.

PSRP

Response pattern by the students belonging to the parent classes of Class c. A more comprehensible arrangement of params.

LCD

Latent Class Distribution. see also plot.exametrika

LFD

Latent Field Distribution. see also plot.exametrika

CMD

Class Membership Distribution.

FRP

Marginal bicluster reference matrix.

FRPIndex

Index of FFP includes the item location parameters B and Beta, the slope parameters A and Alpha, and the monotonicity indices C and Gamma.

TRP

Test Reference Profile

LDPSR

A rearranged set of parameters for output. It includes the field the items contained within that field, and the conditional correct response rate of parent nodes(class) and child node(class).

FieldEstimated

Given vector which correspondence between items and the fields.

Students

Rank Membership Profile matrix.The s-th row vector of \hat{M}_R, \hat{m}_R, is the rank membership profile of Student s, namely the posterior probability distribution representing the student's belonging to the respective latent classes.

NextStage

The next class that easiest for students to move to, its membership probability, class-up odds, and the field required for more.

MG_FitIndices

Multigroup as Null model.See also TestFit

SM_FitIndices

Saturated Model as Null model.See also TestFit

Examples

# Example: Bicluster Network Model (BINET)
# BINET combines Bayesian network model and Biclustering to explore
# local dependency structure among latent classes at each field

# Create field configuration vector based on field assignments
conf <- c(
  1, 5, 5, 5, 9, 9, 6, 6, 6, 6, 2, 7, 7, 11, 11, 7, 7,
  12, 12, 12, 2, 2, 3, 3, 4, 4, 4, 8, 8, 12, 1, 1, 6, 10, 10
)

# Create edge data for network structure between classes
edges_data <- data.frame(
  "From Class (Parent) >>>" = c(
    1, 2, 3, 4, 5, 7, 2, 4, 6, 8, 10, 6, 6, 11, 8, 9, 12
  ),
  ">>> To Class (Child)" = c(
    2, 4, 5, 5, 6, 11, 3, 7, 9, 12, 12, 10, 8, 12, 12, 11, 13
  ),
  "At Field (Locus)" = c(
    1, 2, 2, 3, 4, 4, 5, 5, 5, 5, 5, 7, 8, 8, 9, 9, 12
  )
)

# Save edge data to temporary CSV file
tmp_file <- tempfile(fileext = ".csv")
write.csv(edges_data, file = tmp_file, row.names = FALSE)

# Fit Bicluster Network Model
result.BINET <- BINET(
  J35S515,
  ncls = 13, # Maximum class number from edges (13)
  nfld = 12, # Maximum field number from conf (12)
  conf = conf, # Field configuration vector
  adj_file = tmp_file # Path to the CSV file
)

# Clean up temporary file
unlink(tmp_file)

# Display model results
print(result.BINET)

# Visualize different aspects of the model
plot(result.BINET, type = "Array") # Show bicluster structure
plot(result.BINET, type = "TRP") # Test Response Profile
plot(result.BINET, type = "LRD") # Latent Rank Distribution
plot(result.BINET,
  type = "RMP", # Rank Membership Profiles
  students = 1:9, nc = 3, nr = 3
)
plot(result.BINET,
  type = "FRP", # Field Reference Profiles
  nc = 3, nr = 2
)
plot(result.BINET,
  type = "LDPSR", # Locally Dependent Passing Student Rates
  nc = 3, nr = 2
)

Bayesian Network Model

Description

performs Bayesian Network Model with specified graph structure

Usage

BNM(
  U,
  Z = NULL,
  w = NULL,
  na = NULL,
  g = NULL,
  adj_file = NULL,
  adj_matrix = NULL
)

Arguments

Details

This function performs a Bayesian network analysis on the relationships between items. This corresponds to Chapter 8 of the text. It uses the igraph package for graph visualization and checking the adjacency matrix. You need to provide either a graph object or a CSV file where the graph structure is specified.

Value

nobs

Sample size. The number of rows in the dataset.

testlength

Length of the test. The number of items included in the test.

crr

correct response ratio

TestFitIndices

Overall fit index for the test.See also TestFit

adj

Adjacency matrix

\

param

Learned Parameters

CCRR_table

Correct Response Rate tables

Examples

# Create a Directed Acyclic Graph (DAG) structure for item relationships
# Each row represents a directed edge from one item to another
DAG <-
  matrix(
    c(
      "Item01", "Item02", # Item01 influences Item02
      "Item02", "Item03", # Item02 influences Item03
      "Item02", "Item04", # Item02 influences Item04
      "Item03", "Item05", # Item03 influences Item05
      "Item04", "Item05" # Item04 influences Item05
    ),
    ncol = 2, byrow = TRUE
  )

# Convert the DAG matrix to an igraph object for network analysis
g <- igraph::graph_from_data_frame(DAG)
g

# Create adjacency matrix from the graph
# Shows direct connections between items (1 for connection, 0 for no connection)
adj_mat <- as.matrix(igraph::as_adjacency_matrix(g))
print(adj_mat)

# Fit Bayesian Network Model using the specified adjacency matrix
# Analyzes probabilistic relationships between items based on the graph structure
result.BNM <- BNM(J5S10, adj_matrix = adj_mat)
result.BNM

Biclustering and Ranklustering Analysis

Description

Performs biclustering, ranklustering, or their confirmatory variants on binary response data. These methods simultaneously cluster both examinees and items into homogeneous groups (or ordered ranks for ranklustering). The analysis reveals latent structures and patterns in the data by creating a matrix with rows and columns arranged to highlight block structures.

Usage

Biclustering(
  U,
  ncls = 2,
  nfld = 2,
  Z = NULL,
  w = NULL,
  na = NULL,
  method = "B",
  conf = NULL,
  mic = FALSE,
  maxiter = 100,
  verbose = TRUE
)

Arguments

Details

Biclustering simultaneously clusters both rows (examinees) and columns (items) of a data matrix. Unlike traditional clustering that groups either rows or columns, biclustering identifies submatrices with similar patterns. Ranklustering is a variant that imposes an ordinal structure on the classes, making it suitable for proficiency scaling.

The algorithm uses an Expectation-Maximization approach to iteratively estimate:

1. Field membership of items (which items belong to which fields)

2. Class/rank membership of examinees (which examinees belong to which classes)

3. Field Reference Profiles (probability patterns for each field-class combination)

The confirmatory option allows for pre-specified field assignments, which is useful when there is prior knowledge about item groupings or for testing hypothesized structures.

Value

An object of class "exametrika" and "Biclustering" containing:

model

Model type indicator (1 for biclustering, 2 for ranklustering)

mic

Logical value indicating whether monotonicity constraint was applied

testlength

Number of items in the test

nobs

Number of examinees in the dataset

Nclass

Number of latent classes/ranks specified

Nfield

Number of latent fields specified

N_Cycle

Number of EM iterations performed

LFD

Latent Field Distribution - counts of items assigned to each field

LRD/LCD

Latent Rank/Class Distribution - counts of examinees assigned to each class/rank

FRP

Field Reference Profile matrix - probability of correct response for each field-class combination

FRPIndex

Field Reference Profile indices including location parameters, slope parameters, and monotonicity indices

TRP

Test Reference Profile - expected score for examinees in each class/rank

CMD/RMD

Class/Rank Membership Distribution - sum of membership probabilities across examinees

FieldMembership

Matrix showing the probabilities of each item belonging to each field

ClassMembership

Matrix showing the probabilities of each examinee belonging to each class/rank

SmoothedMembership

Matrix of smoothed class membership probabilities after filtering

FieldEstimated

Vector of the most likely field assignments for each item

ClassEstimated

Vector of the most likely class/rank assignments for each examinee

Students

Data frame containing membership probabilities and classification information for each examinee

FieldAnalysis

Matrix showing field analysis results with item-level information

TestFitIndices

Model fit indices for evaluating the quality of the clustering solution

SOACflg

Logical flag indicating whether Strongly Ordinal Alignment Condition is satisfied

WOACflg

Logical flag indicating whether Weakly Ordinal Alignment Condition is satisfied

References

Shojima, K. (2012). Biclustering of binary data matrices using bilinear models. Behaviormetrika, 39(2), 161-178.

Examples

# Perform Biclustering with Binary method (B)
# Analyze data with 5 fields and 6 classes
result.Bi <- Biclustering(J35S515, nfld = 5, ncls = 6, method = "B")

# Perform Biclustering with Rank method (R)
# Store results for further analysis and visualization
result.Rank <- Biclustering(J35S515, nfld = 5, ncls = 6, method = "R")

# Display the Bicluster Reference Matrix (BRM) as a heatmap
plot(result.Rank, type = "Array")

# Plot Field Reference Profiles (FRP) in a 2x3 grid
# Shows the probability patterns for each field
plot(result.Rank, type = "FRP", nc = 2, nr = 3)

# Plot Rank Membership Profiles (RMP) for students 1-9 in a 3x3 grid
# Shows posterior probability distribution of rank membership
plot(result.Rank, type = "RMP", students = 1:9, nc = 3, nr = 3)

# Example of confirmatory analysis with pre-specified fields
# Assign items 1-10 to field 1, 11-20 to field 2, etc.
field_assignments <- c(rep(1, 10), rep(2, 10), rep(3, 15))
result.Conf <- Biclustering(J35S515, nfld = 3, ncls = 5, conf = field_assignments)

Biserial Correlation

Description

A biserial correlation is a correlation between dichotomous-ordinal and continuous variables.

Usage

BiserialCorrelation(i, t)

Arguments

Value

The biserial correlation coefficient between the two variables.

Binary pattern maker

Description

Binary pattern maker

Usage

BitRespPtn(n)

Arguments

Details

if n <- 1, return 0,1 if n <- 2, return 00,01,10,11 and so on.

Value

binary patterns

Conditional Correct Response Rate

Description

The conditional correct response rate (CCRR) represents the ratio of the students who passed Item C (consequent item) to those who passed Item A (antecedent item). This function is applicable only to binary response data.

Usage

CCRR(U, na = NULL, Z = NULL, w = NULL)

## Default S3 method:
CCRR(U, na = NULL, Z = NULL, w = NULL)

## S3 method for class 'binary'
CCRR(U, na = NULL, Z = NULL, w = NULL)

## S3 method for class 'nominal'
CCRR(U, na = NULL, Z = NULL, w = NULL)

Arguments

Value

A matrix of conditional correct response rates with exametrika class. Each element (i,j) represents the probability of correctly answering item j given that item i was answered correctly.

Note

This function is implemented using a binary data compatibility wrapper and will raise an error if used with polytomous data.

Examples

# example code
# Calculate CCRR using sample dataset J5S10
CCRR(J5S10)

Conditional Selection Rate

Description

Calculate the Conditional Selection Rate (CSR) for polytomous data. CSR measures the proportion of respondents who selected a specific category in item K, given that they selected a particular category in item J.

Usage

CSR(U, na = NULL, Z = NULL, w = NULL)

Arguments

Details

The function returns a nested list structure CSR, where CSR[[j]][[k]] contains a matrix of conditional probabilities. In this matrix, the element at row l and column m represents P(K=m|J=l), which is the probability of selecting category m for item K, given that category l was selected for item J.

Mathematically, for each cell (l,m) in the CSR[[j]][[k]] matrix: CSR[[j]][[k]][l,m] = P(Item K = category m | Item J = category l)

This is calculated as the number of respondents who selected both category l for item J and category m for item K, divided by the total number of respondents who selected category l for item J.

Value

A list of Joint Selection Rate matrices for each item pair.

Examples

# example code
# Calculate CSR using sample dataset J5S1000
CSR(J5S1000)

# Extract the conditional selection rates from item 1 to item 2
csr_1_2 <- CSR(J5S1000)[[1]][[2]]
# This shows the probability of selecting each category in item 2
# given that a specific category was selected in item 1

Classical Test Theory

Description

This function calculates the overall alpha and omega coefficients for the given data matrix. It also computes the alpha coefficient for each item, assuming that item is excluded.

Usage

CTT(U, na = NULL, Z = NULL, w = NULL)

Arguments

Value

Returns a list of class c("exametrika", "CTT") containing two data frames:

Reliability

A data frame with overall reliability coefficients (Alpha and Omega) calculated using different correlation matrices (Covariance, Phi, and Tetrachoric)

ReliabilityExcludingItem

A data frame showing alpha coefficients when each item is excluded, calculated using different correlation matrices

Examples

# using sample dataset
CTT(J15S500)

Dimensionality

Description

The dimensionality is the number of components the test is measuring.

Usage

Dimensionality(U, na = NULL, Z = NULL, w = NULL)

## Default S3 method:
Dimensionality(U, na = NULL, Z = NULL, w = NULL)

## S3 method for class 'binary'
Dimensionality(U, na = NULL, Z = NULL, w = NULL)

## S3 method for class 'rated'
Dimensionality(U, na = NULL, Z = NULL, w = NULL)

## S3 method for class 'ordinal'
Dimensionality(U, na = NULL, Z = NULL, w = NULL)

Arguments

Value

Returns a list of class c("exametrika", "Dimensionality") containing:

Component

Sequence of component numbers

Eigenvalue

Eigenvalues of the tetrachoric correlation matrix

PerOfVar

Percentage of variance explained by each component

CumOfPer

Cumulative percentage of variance explained

Graded Response Model (GRM)

Description

Implements Samejima's (1969) Graded Response Model (GRM), which is an Item Response Theory model for ordered categorical response data. The model estimates discrimination parameters and category threshold parameters for each item. It is widely used in psychological measurement, educational assessment, and other fields that deal with multi-step rating scales.

Usage

GRM(U, na = NULL, Z = NULL, w = NULL, verbose = TRUE)

Arguments

Value

A list of class "exametrika" and "GRM" containing the following elements:

testlength

Length of the test (number of items)

nobs

Sample size (number of rows in the dataset)

params

Matrix containing the estimated item parameters

EAP

Ability parameters of examinees estimated by EAP method

MAP

Ability parameters of examinees estimated by MAP method

PSD

Posterior standard deviation of the ability parameters

ItemFitIndices

Fit indices for each item. See also ItemFit

TestFitIndices

Overall fit indices for the test. See also TestFit

References

Samejima, F. (1969). Estimation of latent ability using a response pattern of graded scores. Psychometrika Monograph Supplement, 34(4, Pt. 2), 1-100.

Examples

# Apply GRM to example data
result <- GRM(J5S1000)
print(result)
plot(result, type = "IRF")
plot(result, type = "IIF")
plot(result, type = "TIF")

IIF for 2PLM

Description

Item Information Function for 2PLM

Usage

IIF2PLM(a, b, theta)

Arguments

Value

Returns a numeric vector representing the item information at each ability level theta. The information is calculated as: I(\theta) = a^2P(\theta)(1-P(\theta))

IIF for 3PLM

Description

Item Information Function for 3PLM

Usage

IIF3PLM(a, b, c, theta)

Arguments

Value

Returns a numeric vector representing the item information at each ability level theta. The information is calculated as: I(\theta) = \frac{a^2(1-P(\theta))(P(\theta)-c)^2}{(1-c)^2P(\theta)}

Infinite Relational Model

Description

The purpose of this method is to find the optimal number of classes C, and optimal number of fields F. It can be found in a single run of the analysis, but it takes a long computation time when the sample size S is large. In addition, this method incorporates the Chinese restaurant process and Gibbs sampling. In detail, See Section 7.8 in Shojima(2022).

Usage

IRM(
  U,
  Z = NULL,
  w = NULL,
  na = NULL,
  gamma_c = 1,
  gamma_f = 1,
  max_iter = 100,
  stable_limit = 5,
  minSize = 20,
  EM_limit = 20,
  seed = 123,
  verbose = TRUE
)

Arguments

Value

nobs

Sample size. The number of rows in the dataset.

testlength

Length of the test. The number of items included in the test.

Nclass

Optimal number of classes.

Nfield

Optimal number of fields.

BRM

Bicluster Reference Matrix

FRP

Field Reference Profile

FRPIndex

Index of FFP includes the item location parameters B and Beta, the slope parameters A and Alpha, and the monotonicity indices C and Gamma.

TRP

Test Reference Profile

FMP

Field Membership Profile

Students

Rank Membership Profile matrix.The s-th row vector of \hat{M}_R, \hat{m}_R, is the rank membership profile of Student s, namely the posterior probability distribution representing the student's belonging to the respective latent classes. It also includes the rank with the maximum estimated membership probability, as well as the rank-up odds and rank-down odds.

LRD

Latent Rank Distribution. see also plot.exametrika

LFD

Latent Field Distribution. see also plot.exametrika

RMD

Rank Membership Distribution.

TestFitIndices

Overall fit index for the test.See also TestFit

Examples

# Fit an Infinite Relational Model (IRM) to determine optimal number of classes and fields
# gamma_c and gamma_f are concentration parameters for the Chinese Restaurant Process
result.IRM <- IRM(J35S515, gamma_c = 1, gamma_f = 1, verbose = TRUE)

# Display the Bicluster Reference Matrix (BRM) as a heatmap
# Shows the discovered clustering structure of items and students
plot(result.IRM, type = "Array")

# Plot Field Reference Profiles (FRP) in a 3-column grid
# Shows the probability patterns for each automatically determined field
plot(result.IRM, type = "FRP", nc = 3)

# Plot Test Reference Profile (TRP)
# Shows the overall response pattern across all fields
plot(result.IRM, type = "TRP")

Estimating Item parameters using EM algorithm

Description

A function for estimating item parameters using the EM algorithm.

Usage

IRT(U, model = 2, na = NULL, Z = NULL, w = NULL, verbose = TRUE)

Arguments

Details

Apply the 2, 3, and 4 parameter logistic models to estimate the item and subject populations. The 4PL model can be described as follows.

P(\theta,a_j,b_j,c_j,d_j)= c_j + \frac{d_j -c_j}{1+exp\{-a_j(\theta - b_j)\}}

a_j, b_j, c_j, and d_j are parameters related to item j, and are parameters that adjust the logistic curve. a_j is called the slope parameter, b_j is the location, c_j is the lower asymptote, and d_j is the upper asymptote parameter. The model includes lower models, and among the 4PL models, the case where d=1 is the 3PL model, and among the 3PL models, the case where c=0 is the 2PL model.

Value

model

number of item parameters you set.

testlength

Length of the test. The number of items included in the test.

nobs

Sample size. The number of rows in the dataset.

params

Matrix containing the estimated item parameters

Q3mat

Q3-matrix developed by Yen(1984)

itemPSD

Posterior standard deviation of the item parameters

ability

Estimated parameters of students ability

ItemFitIndices

Fit index for each item.See also ItemFit

TestFitIndices

Overall fit index for the test.See also TestFit

References

Yen, W. M. (1984) Applied Psychological Measurement, 8, 125-145.

Examples

# Fit a 3-parameter IRT model to the sample dataset
result.IRT <- IRT(J15S500, model = 3)

# Display the first few rows of estimated student abilities
head(result.IRT$ability)

# Plot Item Response Function (IRF) for items 1-6 in a 2x3 grid
plot(result.IRT, type = "IRF", items = 1:6, nc = 2, nr = 3)

# Plot Item Information Function (IIF) for items 1-6 in a 2x3 grid
plot(result.IRT, type = "IIF", items = 1:6, nc = 2, nr = 3)

# Plot the Test Information Function (TIF) for all items
plot(result.IRT, type = "TIF")

Item-Total Biserial Correlation

Description

The Item-Total Biserial Correlation computes the biserial correlation between each item and the total score. This function is applicable only to binary response data.

This correlation provides a measure of item discrimination, indicating how well each item distinguishes between high and low performing examinees.

Usage

ITBiserial(U, na = NULL, Z = NULL, w = NULL)

## Default S3 method:
ITBiserial(U, na = NULL, Z = NULL, w = NULL)

## S3 method for class 'binary'
ITBiserial(U, na = NULL, Z = NULL, w = NULL)

Arguments

Value

A numeric vector of item-total biserial correlations. Values range from -1 to 1, where:

• Values near 1: Strong positive discrimination

• Values near 0: No discrimination

• Negative values: Potential item problems

Note

This function is implemented using a binary data compatibility wrapper and will raise an error if used with polytomous data.

The biserial correlation is generally preferred over the point-biserial correlation when the dichotomization is artificial (i.e., when the underlying trait is continuous).

Examples

# using sample dataset
ITBiserial(J15S500)

Inter-Item Analysis for Psychometric Data

Description

Calculates various relationship metrics between pairs of items in test data. This analysis helps identify item interdependencies, content overlaps, and potential local dependence. For binary data, metrics include joint response rates, conditional probabilities, and several correlation measures. For ordinal/rated data, appropriate correlation measures are calculated.

The following metrics are calculated for binary data:

• JSS: Joint Sample Size - number of examinees responding to both items

• JCRR: Joint Correct Response Rate - proportion of examinees answering both items correctly

• CCRR: Conditional Correct Response Rate - probability of answering one item correctly given a correct response to another item

• IL: Item Lift - ratio of joint correct response rate to the product of marginal rates

• MI: Mutual Information - measure of mutual dependence between items

• Phi: Phi Coefficient - correlation coefficient for binary variables

• Tetrachoric: Tetrachoric Correlation - estimate of Pearson correlation for underlying continuous variables

For ordinal/rated data, the function calculates:

• JSS: Joint Sample Size

• JSR: Joint Selection Rate

• CSR: Conditional Selection Rate

• MI: Mutual Information

• Polychoric: Polychoric Correlation - extension of tetrachoric correlation for ordinal data

Usage

InterItemAnalysis(U, na = NULL, Z = NULL, w = NULL)

Arguments

Details

This function automatically detects the data type and applies appropriate analysis methods:

• For binary data: Calculates tetrachoric correlations and related statistics

• For ordinal/rated data: Calculates polychoric correlations and related statistics

• For nominal data: Returns an error (not supported)

Inter-item analysis is useful for:

• Identifying groups of highly related items

• Detecting local dependence between items

• Evaluating test dimensionality

• Informing item selection and test construction

Value

For binary data, an object of class "exametrika" and "IIAnalysis" containing:

JSS

Joint Sample Size matrix - N(i,j) shows number of examinees who responded to both items i and j

JCRR

Joint Correct Response Rate matrix - P(Xi=1, Xj=1) shows probability of correct responses to both items

CCRR

Conditional Correct Response Rate matrix - P(Xi=1|Xj=1) shows probability of correct response to item i given correct response to item j

IL

Item Lift matrix - P(Xi=1, Xj=1)/(P(Xi=1)*P(Xj=1)) measures association strength

MI

Mutual Information matrix - measures information shared between items

Phi

Phi Coefficient matrix - correlation coefficient between binary variables

Tetrachoric

Tetrachoric Correlation matrix - correlation between underlying continuous variables

For ordinal/rated data, an object of class "exametrika" and "IIAnalysis.ordinal" containing:

JSS

Joint Sample Size matrix

JSR

Joint Selection Rate matrix - frequencies of joint category selections

CSR

Conditional Selection Rate matrix - probabilities of response categories conditional on other items

MI

Mutual Information matrix

Polychoric

Polychoric Correlation matrix - correlations between underlying continuous variables

See Also

dataFormat for data preparation, CTT for Classical Test Theory analysis

Examples

# Basic usage with binary data
ii_analysis <- InterItemAnalysis(J15S500)

# View joint sample sizes
head(ii_analysis$JSS)

# View tetrachoric correlations
head(ii_analysis$Tetrachoric)

# Find pairs of items with high mutual information (potential local dependence)
high_MI <- which(ii_analysis$MI > 0.2 & upper.tri(ii_analysis$MI), arr.ind = TRUE)
if (nrow(high_MI) > 0) {
  print("Item pairs with high mutual information:")
  print(high_MI)
}

# Example with ordinal data
ordinal_analysis <- InterItemAnalysis(J15S3810)

# View polychoric correlations for ordinal data
head(ordinal_analysis$Polychoric)

Item Entropy

Description

The item entropy is an indicator of the variability or randomness of the responses. This function is applicable only to binary response data.

The entropy value represents the uncertainty or information content of the response pattern for each item, measured in bits. Maximum entropy (1 bit) occurs when correct and incorrect responses are equally likely (p = 0.5).

Usage

ItemEntropy(U, na = NULL, Z = NULL, w = NULL)

## Default S3 method:
ItemEntropy(U, na = NULL, Z = NULL, w = NULL)

## S3 method for class 'binary'
ItemEntropy(U, na = NULL, Z = NULL, w = NULL)

## S3 method for class 'ordinal'
ItemEntropy(U, na = NULL, Z = NULL, w = NULL)

Arguments

Details

The item entropy is calculated as:

e_j = -p_j\log_2p_j-(1-p_j)\log_2(1-p_j)

where p_j is the correct response rate for item j.

The entropy value has the following properties:

• Maximum value of 1 bit when p = 0.5 (most uncertainty)

• Minimum value of 0 bits when p = 0 or 1 (no uncertainty)

• Higher values indicate more balanced response patterns

• Lower values indicate more predictable response patterns

Value

A numeric vector of entropy values for each item, measured in bits. Values range from 0 to 1, where:

• 1: maximum uncertainty (p = 0.5)

• 0: complete certainty (p = 0 or 1)

• Values near 1 indicate items with balanced response patterns

• Values near 0 indicate items with extreme response patterns

Note

This function is implemented using a binary data compatibility wrapper and will raise an error if used with polytomous data.

Examples

# using sample dataset
ItemEntropy(J5S10)

Model Fit Functions for Items

Description

A general function that returns the model fit indices.

Usage

ItemFit(U, Z, ell_A, nparam)

Arguments

Value

model_log_like

log likelihood of analysis model

bench_log_like

log likelihood of benchmark model

null_log_like

log likelihood of null model

model_Chi_sq

Chi-Square statistics for analysis model

null_Chi_sq

Chi-Square statistics for null model

model_df

degrees of freedom of analysis model

null_df

degrees of freedom of null model

NFI

Normed Fit Index. Lager values closer to 1.0 indicate a better fit.

RFI

Relative Fit Index. Lager values closer to 1.0 indicate a better fit.

IFI

Incremental Fit Index. Lager values closer to 1.0 indicate a better fit.

TLI

Tucker-Lewis Index. Lager values closer to 1.0 indicate a better fit.

CFI

Comparative Fit Index. Lager values closer to 1.0 indicate a better fit.

RMSEA

Root Mean Square Error of Approximation. Smaller values closer to 0.0 indicate a better fit.

AIC

Akaike Information Criterion. A lower value indicates a better fit.

CAIC

Consistent AIC.A lower value indicates a better fit.

BIC

Bayesian Information Criterion. A lower value indicates a better fit.

IIF for 4PLM

Description

Item Information Function for 4PLM

Usage

ItemInformationFunc(a = 1, b, c = 0, d = 1, theta)

Arguments

Value

Returns a numeric vector representing the item information at each ability level theta. The information is calculated based on the first derivative of the log-likelihood of the 4PL model with respect to theta.

Item Lift

Description

The lift is a commonly used index in a POS data analysis. The item lift of Item k to Item j is defined as follow: l_{jk} = \frac{p_{k\mid j}}{p_k} This function is applicable only to binary response data.

Usage

ItemLift(U, na = NULL, Z = NULL, w = NULL)

## Default S3 method:
ItemLift(U, na = NULL, Z = NULL, w = NULL)

## S3 method for class 'binary'
ItemLift(U, na = NULL, Z = NULL, w = NULL)

Arguments

Value

A matrix of item lift values with exametrika class. Each element (j,k) represents the lift value of item k given item j, which indicates how much more likely item k is to be correct given that item j was answered correctly.

Note

This function is implemented using a binary data compatibility wrapper and will raise an error if used with polytomous data.

References

Brin, S., Motwani, R., Ullman, J., & Tsur, S. (1997). Dynamic itemset counting and implication rules for market basket data. In Proceedings of ACM SIGMOD International Conference on Management of Data (pp. 255–264). https://dl.acm.org/doi/10.1145/253262.253325

Examples

# example code
# Calculate ItemLift using sample dataset J5S10
ItemLift(J5S10)

Item Odds

Description

Item Odds are defined as the ratio of Correct Response Rate to Incorrect Response Rate:

O_j = \frac{p_j}{1-p_j}

where p_j is the correct response rate for item j. This function is applicable only to binary response data.

The odds value represents how many times more likely a correct response is compared to an incorrect response. For example, an odds of 2 means students are twice as likely to answer correctly as incorrectly.

Usage

ItemOdds(U, na = NULL, Z = NULL, w = NULL)

## Default S3 method:
ItemOdds(U, na = NULL, Z = NULL, w = NULL)

## S3 method for class 'binary'
ItemOdds(U, na = NULL, Z = NULL, w = NULL)

Arguments

Value

A numeric vector of odds values for each item. Values range from 0 to infinity, where:

• odds > 1: correct response more likely than incorrect

• odds = 1: equally likely

• odds < 1: incorrect response more likely than correct

Note

This function is implemented using a binary data compatibility wrapper and will raise an error if used with polytomous data.

Examples

# using sample dataset
ItemOdds(J5S10)

Generate Item Report for Non-Binary Test Data

Description

Calculates item-level statistics for non-binary test data, including response rates, basic descriptive statistics, and item-total correlations.

Usage

ItemReport(U, na = NULL, Z = NULL, w = NULL)

Arguments

Details

This function is intended for non-binary (ordinal or rated) response data. It provides detailed statistics for each item in the test, focusing on response patterns and the relationship between individual items and overall test performance. If binary data is provided, an error message will be displayed.

Value

An object of class "exametrika" and "QitemStatistics" containing:

ItemLabel

Labels identifying each item

Obs

Number of valid responses for each item

ObsRatio

Proportion of valid responses for each item (range: 0-1)

ItemMean

Mean score of each item

ItemSD

Standard deviation of each item score

ItemCORR

Item-total correlation coefficients - correlation between item scores and total test scores

ItemCORR_R

Corrected item-total correlation coefficients - correlation between item scores and total test scores excluding the target item

Examples

# Generate item report for sample ordinal data
item_stats <- ItemReport(J15S3810)

# View first few rows of the item report
head(item_stats)

# Example with rated data including custom missing value indicator
item_stats2 <- ItemReport(J35S5000, na = -99)

Simple Item Statistics

Description

This function calculates statistics for each item, with different metrics available depending on the data type (binary, ordinal, or rated).

Usage

ItemStatistics(U, na = NULL, Z = NULL, w = NULL)

## Default S3 method:
ItemStatistics(U, na = NULL, Z = NULL, w = NULL)

## S3 method for class 'binary'
ItemStatistics(U, na = NULL, Z = NULL, w = NULL)

## S3 method for class 'ordinal'
ItemStatistics(U, na = NULL, Z = NULL, w = NULL)

Arguments

Value

For binary data:

ItemLabel

Label identifying each item

NR

Number of Respondents for each item

CRR

Correct Response Rate denoted as $p_j$.

ODDs

Item Odds is the ratio of the correct response rate to the incorrect response rate. Defined as o_j = \frac{p_j}{1-p_j}

Threshold

Item Threshold is a measure of difficulty based on a standard normal distribution.

Entropy

Item Entropy is an indicator of the variability or randomness of the responses. Defined as e_j=-p_j \log_2 p_j - (1-p_j)\log_2(1-p_j)

ITCrr

Item-total Correlation is a Pearson's correlation of an item with the Number-Right score.

For ordinal polytomous data:

ItemLabel

Label identifying each item

NR

Number of Respondents for each item

Threshold

Matrix of threshold values for each item's category boundaries, based on a standard normal distribution. For an item with K categories, there are K-1 thresholds.

Entropy

Item Entropy calculated using the category probabilities. Unlike binary data, this is calculated using the formula e_j = -\sum_{k=1}^{K_j} p_{jk} \log_{K_j} p_{jk}, where K_j is the number of categories for item j.

ITCrr

Item-total Correlation calculated using polyserial correlation, which accounts for the ordinal nature of the item responses and the continuous total score.

Note

For rated data, the function processes the data as binary, with each response being compared to the correct answer to determine correctness.

Examples

# using sample dataset(binary)
ItemStatistics(J15S500)

Item Threshold

Description

Item threshold is a measure of difficulty based on a standard normal distribution. This function is applicable only to binary response data.

The threshold is calculated as:

\tau_j = \Phi^{-1}(1-p_j)

where \Phi^{-1} is the inverse standard normal distribution function and p_j is the correct response rate for item j.

Higher threshold values indicate more difficult items, as they represent the point on the standard normal scale above which examinees tend to answer incorrectly.

Usage

ItemThreshold(U, na = NULL, Z = NULL, w = NULL)

## S3 method for class 'binary'
ItemThreshold(U, na = NULL, Z = NULL, w = NULL)

## S3 method for class 'ordinal'
ItemThreshold(U, na = NULL, Z = NULL, w = NULL)

Arguments

Value

A numeric vector of threshold values for each item on the standard normal scale. Typical values range from about -3 to 3, where:

• Positive values indicate difficult items

• Zero indicates items of medium difficulty (50% correct)

• Negative values indicate easy items

Note

This function is implemented using a binary data compatibility wrapper and will raise an error if used with polytomous data.

Examples

# using sample dataset
ItemThreshold(J5S10)

Item-Total Correlation

Description

Item-Total correlation (ITC) is a Pearson's correlation of an item with the Number-Right Score (NRS) or total score. This function is applicable only to binary response data.

The ITC is a measure of item discrimination, indicating how well an item distinguishes between high and low performing examinees.

Usage

ItemTotalCorr(U, na = NULL, Z = NULL, w = NULL)

## Default S3 method:
ItemTotalCorr(U, na = NULL, Z = NULL, w = NULL)

## S3 method for class 'binary'
ItemTotalCorr(U, na = NULL, Z = NULL, w = NULL)

## S3 method for class 'ordinal'
ItemTotalCorr(U, na = NULL, Z = NULL, w = NULL)

Arguments

Details

The correlation is calculated between:

• Each item's responses (0 or 1)

• The total test score (sum of correct responses)

Higher positive correlations indicate items that better discriminate between high and low ability examinees.

Value

A numeric vector of item-total correlations. Values typically range from -1 to 1, where:

• Values near 1: Strong positive discrimination

• Values near 0: No discrimination

• Negative values: Potential item problems (lower ability students performing better than higher ability students)

Note

This function is implemented using a binary data compatibility wrapper and will raise an error if used with polytomous data.

Values below 0.2 might indicate problematic items that should be reviewed. Values above 0.3 are generally considered acceptable.

Examples

# using sample dataset
ItemTotalCorr(J15S500)

J12S5000

Description

A binary response dataset for test analysis

Usage

J12S5000

Format

An exametrika class object with 5000 students and 12 items containing binary (0/1) responses

Source

http://sh0j1ma.stars.ne.jp/exmk/

J15S3810

Description

A ordinal response dataset for test analysis

Usage

J15S3810

Format

An exametrika class object with 3810 students and 15 items containing nominal responses with 4 categories

J15S500

Description

A binary response dataset for test analysis

Usage

J15S500

Format

An exametrika class object with 500 students and 15 items containing binary (0/1) responses

Source

http://sh0j1ma.stars.ne.jp/exmk/

J20S400

Description

A binary response dataset for test analysis

Usage

J20S400

Format

An exametrika class object with 400 students and 20 items containing binary (0/1) responses

Source

http://sh0j1ma.stars.ne.jp/exmk/

J35S5000

Description

A rated response dataset for test analysis

Usage

J35S5000

Format

An exametrika class object with 5000 students and 35 items containing polytomous responses with correct answers

J35S515

Description

A binary response dataset for test analysis

Usage

J35S515

Format

An exametrika class object with 515 students and 35 items containing binary (0/1) responses

Source

http://sh0j1ma.stars.ne.jp/exmk/

J50S100

Description

A simulated binary dataset for test analysis. This is a synthetic dataset generated using random number generation for demonstration and testing purposes.

Usage

J50S100

Format

An exametrika class object with 100 students and 50 items containing binary responses

J5S10

Description

A binary response dataset for test analysis

Usage

J5S10

Format

An exametrika class object with 5 students and 10 items containing binary (0/1) responses

Source

http://sh0j1ma.stars.ne.jp/exmk/

J5S1000

Description

A simulated ordinal dataset for test analysis. This is a synthetic dataset generated using random number generation for demonstration and testing purposes.

Usage

J5S1000

Format

An exametrika class object with 1000 students and 5 items containing ordinal responses

Joint Correct Response Rate

Description

The joint correct response rate (JCRR) is the rate of students who passed both items. This function is applicable only to binary response data. For non-binary data, it will automatically redirect to the JSR function with an appropriate message.

Usage

JCRR(U, na = NULL, Z = NULL, w = NULL)

## Default S3 method:
JCRR(U, na = NULL, Z = NULL, w = NULL)

## S3 method for class 'binary'
JCRR(U, na = NULL, Z = NULL, w = NULL)

## S3 method for class 'nominal'
JCRR(U, na = NULL, Z = NULL, w = NULL)

Arguments

Value

A matrix of joint correct response rates with exametrika class. Each element (i,j) represents the proportion of students who correctly answered both items i and j.

Note

This function is implemented using a binary data compatibility wrapper and will raise an error if used with polytomous data.

Examples

# example code
# Calculate JCRR using sample dataset J5S10
JCRR(J5S10)

Joint Selection Rate

Description

Calculate the Joint Selection Rate (JSR) for polytomous data. JSR measures the proportion of respondents who selected specific category combinations between pairs of items. For each pair of items (j,k), it returns a contingency table showing the joint probability of selecting each category combination.

Usage

JSR(U, na = NULL, Z = NULL, w = NULL)

Arguments

Value

A list of Joint Selection Rate matrices for each item pair.

Examples

# example code
# Calculate JCRR using sample dataset J5S1000
JSR(J5S1000)

Jacobian function for GRM

Description

Jacobian function for GRM

Usage

Jacobian_grm(target, nitems, ncat)

Arguments

Joint Sample Size

Description

The joint sample size is a matrix whose elements are the number of individuals who responded to each pair of items.

Usage

JointSampleSize(U, na = NULL, Z = NULL, w = NULL)

## Default S3 method:
JointSampleSize(U, na = NULL, Z = NULL, w = NULL)

## S3 method for class 'binary'
JointSampleSize(U, na = NULL, Z = NULL, w = NULL)

Arguments

Value

Returns a matrix of class c("exametrika", "matrix") where each element (i,j) represents the number of students who responded to both item i and item j. The diagonal elements represent the total number of responses for each item.

Note

This function is implemented using a binary data compatibility wrapper and will raise an error if used with polytomous data.

Latent Class Analysis

Description

Performs Latent Class Analysis (LCA) on binary response data using the Expectation-Maximization (EM) algorithm. LCA identifies unobserved (latent) subgroups of examinees with similar response patterns, and estimates both the class characteristics and individual membership probabilities.

Usage

LCA(U, ncls = 2, na = NULL, Z = NULL, w = NULL, maxiter = 100)

Arguments

Details

Latent Class Analysis is a statistical method for identifying unobserved subgroups within a population based on observed response patterns. It assumes that examinees belong to one of several distinct latent classes, and that the probability of a correct response to each item depends on class membership.

The algorithm proceeds by:

1. Initializing class reference probabilities

2. Computing posterior class membership probabilities for each examinee (E-step)

3. Re-estimating class reference probabilities based on these memberships (M-step)

4. Iterating until convergence or reaching the maximum number of iterations

Unlike Item Response Theory (IRT), LCA treats latent variables as categorical rather than continuous, identifying distinct profiles rather than positions on a continuum.

Value

An object of class "exametrika" and "LCA" containing:

testlength

Length of the test (number of items).

nobs

Sample size (number of rows in the dataset).

Nclass

Number of latent classes specified.

N_Cycle

Number of EM algorithm iterations performed.

TRP

Test Reference Profile vector showing expected scores for each latent class. Calculated as the column sum of the estimated class reference matrix.

LCD

Latent Class Distribution vector showing the number of examinees assigned to each latent class.

CMD

Class Membership Distribution vector showing the sum of membership probabilities for each latent class.

Students

Class Membership Profile matrix showing the posterior probability of each examinee belonging to each latent class. The last column ("Estimate") indicates the most likely class assignment.

IRP

Item Reference Profile matrix where each row represents an item and each column represents a latent class. Values indicate the probability of a correct response for members of that class.

ItemFitIndices

Fit indices for each item. See also ItemFit.

TestFitIndices

Overall fit indices for the test. See also TestFit.

References

Goodman, L. A. (1974). Exploratory latent structure analysis using both identifiable and unidentifiable models. Biometrika, 61(2), 215-231.

Lazarsfeld, P. F., & Henry, N. W. (1968). Latent structure analysis. Boston: Houghton Mifflin.

Examples

# Fit a Latent Class Analysis model with 5 classes to the sample dataset
result.LCA <- LCA(J15S500, ncls = 5)

# Display the first few rows of student class membership probabilities
head(result.LCA$Students)

# Plot Item Response Profiles (IRP) for items 1-6 in a 2x3 grid
# Shows probability of correct response for each item across classes
plot(result.LCA, type = "IRP", items = 1:6, nc = 2, nr = 3)

# Plot Class Membership Probabilities (CMP) for students 1-9 in a 3x3 grid
# Shows probability distribution of class membership for each student
plot(result.LCA, type = "CMP", students = 1:9, nc = 3, nr = 3)

# Plot Test Response Profile (TRP) showing expected scores for each class
plot(result.LCA, type = "TRP")

# Plot Latent Class Distribution (LCD) showing class sizes
plot(result.LCA, type = "LCD")

# Compare models with different numbers of classes
# (In practice, you might try more class counts)
lca2 <- LCA(J15S500, ncls = 2)
lca3 <- LCA(J15S500, ncls = 3)
lca4 <- LCA(J15S500, ncls = 4)
lca5 <- LCA(J15S500, ncls = 5)

# Compare BIC values to select optimal number of classes
# (Lower BIC indicates better fit)
data.frame(
  Classes = 2:5,
  BIC = c(
    lca2$TestFitIndices$BIC,
    lca3$TestFitIndices$BIC,
    lca4$TestFitIndices$BIC,
    lca5$TestFitIndices$BIC
  )
)

Local Dependence Biclustering

Description

Latent dependence Biclustering, which incorporates biclustering and a Bayesian network model.

Usage

LDB(
  U,
  Z = NULL,
  w = NULL,
  na = NULL,
  ncls = 2,
  method = "R",
  conf = NULL,
  g_list = NULL,
  adj_list = NULL,
  adj_file = NULL,
  verbose = FALSE
)

Arguments

Value

nobs

Sample size. The number of rows in the dataset.

testlength

Length of the test. The number of items included in the test.

Nclass

Optimal number of classes.

Nfield

Optimal number of fields.

crr

Correct Response Rate

ItemLabel

Label of Items

FieldLabel

Label of Fields

adj_list

List of Adjacency matrix used in the model

g_list

List of graph object used in the model

IRP

List of Estimated Parameters. This object is three-dimensional PIRP array, where each dimension represents the number of rank,number of field, and Dmax. Dmax denotes the maximum number of correct response patterns for each field.

LFD

Latent Field Distribution. see also plot.exametrika

LRD

Latent Rank Distribution. see also plot.exametrika

FRP

Marginal Field Reference Matrix

FRPIndex

Index of FFP includes the item location parameters B and Beta, the slope parameters A and Alpha, and the monotonicity indices C and Gamma.

CCRR_table

This table is a rearrangement of IRP into a data.frame format for output, consisting of combinations of rank ,field and PIRP.

TRP

Test Reference Profile

RMD

Rank Membership Distribution.

FieldEstimated

Given vector which correspondence between items and the fields.

ClassEstimated

An index indicating which class a student belongs to, estimated by confirmatory Ranklustering.

Students

Rank Membership Profile matrix.The s-th row vector of \hat{M}_R, \hat{m}_R, is the rank membership profile of Student s, namely the posterior probability distribution representing the student's belonging to the respective latent classes. It also includes the rank with the maximum estimated membership probability, as well as the rank-up odds and rank-down odds.

TestFitIndices

Overall fit index for the test.See also TestFit

Examples

# Example: Latent Dirichlet Bayesian Network model
# Create field configuration vector based on field assignments
conf <- c(
  1, 6, 6, 8, 9, 9, 4, 7, 7, 7, 5, 8, 9, 10, 10, 9, 9,
  10, 10, 10, 2, 2, 3, 3, 5, 5, 6, 9, 9, 10, 1, 1, 7, 9, 10
)

# Create edge data for the network structure between fields
edges_data <- data.frame(
  "From Field (Parent) >>>" = c(
    6, 4, 5, 1, 1, 4, # Class/Rank 2
    3, 4, 6, 2, 4, 4, # Class/Rank 3
    3, 6, 4, 1, # Class/Rank 4
    7, 9, 6, 7 # Class/Rank 5
  ),
  ">>> To Field (Child)" = c(
    8, 7, 8, 7, 2, 5, # Class/Rank 2
    5, 8, 8, 4, 6, 7, # Class/Rank 3
    5, 8, 5, 8, # Class/Rank 4
    10, 10, 8, 9 # Class/Rank 5
  ),
  "At Class/Rank (Locus)" = c(
    2, 2, 2, 2, 2, 2, # Class/Rank 2
    3, 3, 3, 3, 3, 3, # Class/Rank 3
    4, 4, 4, 4, # Class/Rank 4
    5, 5, 5, 5 # Class/Rank 5
  )
)

# Save edge data to temporary CSV file
tmp_file <- tempfile(fileext = ".csv")
write.csv(edges_data, file = tmp_file, row.names = FALSE)

# Fit Latent Dirichlet Bayesian Network model
result.LDB <- LDB(
  U = J35S515,
  ncls = 5, # Number of latent classes
  conf = conf, # Field configuration vector
  adj_file = tmp_file # Path to the CSV file
)

# Clean up temporary file
unlink(tmp_file)

# Display model results
print(result.LDB)

# Visualize different aspects of the model
plot(result.LDB, type = "Array") # Show bicluster structure
plot(result.LDB, type = "TRP") # Test Response Profile
plot(result.LDB, type = "LRD") # Latent Rank Distribution
plot(result.LDB,
  type = "RMP", # Rank Membership Profiles
  students = 1:9, nc = 3, nr = 3
)
plot(result.LDB,
  type = "FRP", # Field Reference Profiles
  nc = 3, nr = 2
)
# Field PIRP Profile showing correct answer counts for each rank and field
plot(result.LDB, type = "FieldPIRP")

Local Dependence Latent Rank Analysis

Description

performs local dependence latent lank analysis(LD_LRA) by Shojima(2011)

Usage

LDLRA(
  U,
  Z = NULL,
  w = NULL,
  na = NULL,
  ncls = 2,
  method = "R",
  g_list = NULL,
  adj_list = NULL,
  adj_file = NULL,
  verbose = FALSE
)

Arguments

Details

This function is intended to perform LD-LRA. LD-LRA is an analysis that combines LRA and BNM, and it is used to analyze the network structure among items in the latent rank. In this function, structural learning is not performed, so you need to provide item graphs for each rank as separate files. The file format for this is plain text CSV that includes edges (From, To) and rank numbers.

Value

nobs

Sample size. The number of rows in the dataset.

testlength

Length of the test. The number of items included in the test.

crr

correct response ratio

adj_list

adjacency matrix list

g_list

graph list

referenceMatrix

Learned Parameters.A three-dimensional array of patterns where item x rank x pattern.

IRP

Marginal Item Reference Matrix

IRPIndex

IRP Indices which include Alpha, Beta, Gamma.

TRP

Test Reference Profile matrix.

LRD

latent Rank/Class Distribution

RMD

Rank/Class Membership Distribution

TestFitIndices

Overall fit index for the test.See also TestFit

Estimation_table

Estimated parameters tables.

CCRR_table

Correct Response Rate tables

Studens

Student information. It includes estimated class membership, probability of class membership, RUO, and RDO.

Examples

# Create sample DAG structure with different rank levels
# Format: From, To, Rank
DAG_dat <- matrix(c(
  "From", "To", "Rank",
  "Item01", "Item02", "1", # Simple structure for Rank 1
  "Item01", "Item02", "2", # More complex structure for Rank 2
  "Item02", "Item03", "2",
  "Item01", "Item02", "3", # Additional connections for Rank 3
  "Item02", "Item03", "3",
  "Item03", "Item04", "3"
), ncol = 3, byrow = TRUE)

# Method 1: Directly use graph and adjacency lists
g_list <- list()
adj_list <- list()

for (i in 1:3) {
  adj_R <- DAG_dat[DAG_dat[, 3] == as.character(i), 1:2, drop = FALSE]
  g_tmp <- igraph::graph_from_data_frame(
    d = data.frame(
      From = adj_R[, 1],
      To = adj_R[, 2]
    ),
    directed = TRUE
  )
  adj_tmp <- igraph::as_adjacency_matrix(g_tmp)
  g_list[[i]] <- g_tmp
  adj_list[[i]] <- adj_tmp
}

# Fit Local Dependence Latent Rank Analysis
result.LDLRA1 <- LDLRA(J12S5000,
  ncls = 3,
  g_list = g_list,
  adj_list = adj_list
)

# Plot Item Reference Profiles (IRP) in a 4x3 grid
# Shows the probability patterns of correct responses for each item across ranks
plot(result.LDLRA1, type = "IRP", nc = 4, nr = 3)

# Plot Test Reference Profile (TRP)
# Displays the overall pattern of correct response probabilities across ranks
plot(result.LDLRA1, type = "TRP")

# Plot Latent Rank Distribution (LRD)
# Shows the distribution of students across different ranks
plot(result.LDLRA1, type = "LRD")

LDparam set

Description

A function that extracts only the estimation of graph parameters after the rank estimation is completed.

Usage

LD_param_est(tmp, adj_list, classRefMat, ncls, smoothpost)

Arguments

Latent Rank Analysis

Description

A general function for estimating Latent Rank Analysis across different response types. This function automatically dispatches to the appropriate method based on the response type:

• For binary data (LRA.binary): Analysis using either SOM or GTM method

• For ordinal data (LRA.ordinal): Analysis using the GTM method with category thresholds

• For rated data (LRA.rated): Analysis using the GTM method with rating categories

Latent Rank Analysis identifies underlying rank structures in test data and assigns examinees to these ranks based on their response patterns.

Usage

LRA(U, ...)

## Default S3 method:
LRA(U, na = NULL, Z = Z, w = w, ...)

## S3 method for class 'binary'
LRA(
  U,
  nrank = 2,
  method = "GTM",
  mic = FALSE,
  maxiter = 100,
  BIC.check = FALSE,
  seed = NULL,
  verbose = TRUE,
  ...
)

## S3 method for class 'ordinal'
LRA(
  U,
  nrank = 2,
  mic = FALSE,
  maxiter = 100,
  trapezoidal = 0,
  eps = 1e-04,
  verbose = TRUE,
  ...
)

## S3 method for class 'rated'
LRA(
  U,
  nrank = 2,
  mic = FALSE,
  maxiter = 100,
  trapezoidal = 0,
  eps = 1e-04,
  minFreqRatio = 0,
  verbose = TRUE,
  ...
)

Arguments

Value

A list of class "exametrika" and the specific subclass (e.g., "LRA", "LRAordinal", "LRArated") containing the following common elements:

testlength

Length of the test (number of items).

nobs

Sample size (number of rows in the dataset).

Nrank

Number of latent ranks specified.

N_Cycle

Number of EM algorithm iterations performed.

TRP

Test Reference Profile vector showing expected scores at each rank.

LRD

Latent Rank Distribution vector showing the number of examinees at each rank.

RMD

Rank Membership Distribution vector showing the sum of probabilities for each rank.

Students

Rank Membership Profile matrix showing the posterior probabilities of examinees belonging to each rank, along with their estimated ranks and odds ratios.

ItemFitIndices

Fit indices for each item. See also ItemFit.

TestFitIndices

Overall fit indices for the test. See also TestFit.

Each subclass returns additional specific elements, detailed in their respective documentation.

For binary data (LRA.binary), the returned list additionally includes:

IRP

Item Reference Profile matrix showing the probability of correct response for each item across different ranks.

IRPIndex

Item Response Profile indices including the location parameters B and Beta, slope parameters A and Alpha, and monotonicity indices C and Gamma.

For ordinal data (LRA.ordinal), the returned list additionally includes:

ScoreReport

Descriptive statistics of test performance, including sample size, test length, central tendency, variability, distribution characteristics, and reliability.

ItemReport

Basic statistics for each item including category proportions and item-total correlations.

ICBR

Item Category Boundary Reference matrix showing cumulative probabilities for rank-category combinations.

ICRP

Item Category Reference Profile matrix showing probability of response in each category by rank.

ScoreRankCorr

Spearman's correlation between test scores and estimated ranks.

RankQuantCorr

Spearman's correlation between estimated ranks and quantile groups.

ScoreRank

Contingency table of raw scores by estimated ranks.

ScoreMembership

Expected rank memberships for each raw score.

RankQuantile

Cross-tabulation of rank frequencies and quantile groups.

MembQuantile

Cross-tabulation of rank membership probabilities and quantile groups.

CatQuant

Response patterns across item categories and quantile groups.

For rated data (LRA.rated), the returned list additionally includes:

ScoreReport

Descriptive statistics of test performance, including sample size, test length, central tendency, variability, distribution characteristics, and reliability.

ItemReport

Basic statistics for each item including category proportions and item-total correlations.

ICRP

Item Category Reference Profile matrix showing probability of response in each category by rank.

ScoreRankCorr

Spearman's correlation between test scores and estimated ranks.

RankQuantCorr

Spearman's correlation between estimated ranks and quantile groups.

ScoreRank

Contingency table of raw scores by estimated ranks.

ScoreMembership

Expected rank memberships for each raw score.

RankQuantile

Cross-tabulation of rank frequencies and quantile groups.

MembQuantile

Cross-tabulation of rank membership probabilities and quantile groups.

ItemQuantileRef

Reference values for each item across quantile groups.

CatQuant

Response patterns across item categories and quantile groups.

Binary Data Method

LRA.binary analyzes dichotomous (0/1) response data using either Self-Organizing Maps (SOM) or Gaussian Topographic Mapping (GTM).

Ordinal Data Method

LRA.ordinal analyzes ordered categorical data with multiple thresholds, such as Likert-scale responses or graded items.

Rated Data Method

LRA.rated analyzes data with ratings assigned to each response, such as partially-credited items or preference scales where response categories have different weights.

See Also

plot.exametrika for visualizing LRA results.

Examples

# Binary data example
# Fit a Latent Rank Analysis model with 6 ranks to binary data
result.LRA <- LRA(J15S500, nrank = 6)

# Display the first few rows of student rank membership profiles
head(result.LRA$Students)

# Plot Item Reference Profiles (IRP) for the first 6 items
plot(result.LRA, type = "IRP", items = 1:6, nc = 2, nr = 3)

# Plot Test Reference Profile (TRP) showing expected scores at each rank
plot(result.LRA, type = "TRP")



# Ordinal data example
# Fit a Latent Rank Analysis model with 3 ranks to ordinal data
result.LRAord <- LRA(J15S3810, nrank = 3, mic = TRUE)

# Plot score distributions
plot(result.LRAord, type = "ScoreFreq")
plot(result.LRAord, type = "ScoreRank")

# Plot category response patterns for items 1-6
plot(result.LRAord, type = "ICBR", items = 1:6, nc = 3, nr = 2)
plot(result.LRAord, type = "ICRP", items = 1:6, nc = 3, nr = 2)



# Rated data example
# Fit a Latent Rank Analysis model with 10 ranks to rated data
result.LRArated <- LRA(J35S5000, nrank = 10, mic = TRUE)

# Plot score distributions
plot(result.LRArated, type = "ScoreFreq")
plot(result.LRArated, type = "ScoreRank")

# Plot category response patterns for items 1-6
plot(result.LRArated, type = "ICRP", items = 1:6, nc = 3, nr = 2)

Four-Parameter Logistic Model

Description

The four-parameter logistic model is a model where one additional parameter d, called the upper asymptote parameter, is added to the 3PLM.

Usage

LogisticModel(a = 1, b, c = 0, d = 1, theta)

Arguments

Value

Returns a numeric vector of probabilities between c and d, representing the probability of a correct response given the ability level theta. The probability is calculated using the formula: P(\theta) = c + \frac{(d-c)}{1 + e^{-a(\theta-b)}}

Mutual Information

Description

Mutual Information is a measure that represents the degree of interdependence between two items. This function is applicable to both binary and polytomous response data. The measure is calculated using the joint probability distribution of responses between item pairs and their marginal probabilities.

Usage

MutualInformation(U, na = NULL, Z = NULL, w = NULL, base = 2)

## Default S3 method:
MutualInformation(U, na = NULL, Z = NULL, w = NULL, base = 2)

## S3 method for class 'binary'
MutualInformation(U, na = NULL, Z = NULL, w = NULL, base = 2)

## S3 method for class 'ordinal'
MutualInformation(U, na = NULL, Z = NULL, w = NULL, base = 2)

Arguments

Details

For binary data, the following formula is used:

MI_{jk} = p_{00} \log_2 \frac{p_{00}}{(1-p_j)(1-p_k)} + p_{01} \log_2 \frac{p_{01}}{(1-p_j)p_k} + p_{10} \log_2 \frac{p_{10}}{p_j(1-p_k)} + p_{11} \log_2 \frac{p_{11}}{p_jp_k}

Where:

• p_{00} is the joint probability of incorrect responses to both items j and k

• p_{01} is the joint probability of incorrect response to item j and correct to item k

• p_{10} is the joint probability of correct response to item j and incorrect to item k

• p_{11} is the joint probability of correct responses to both items j and k

For polytomous data, the following formula is used:

MI_{jk} = \sum_{j=1}^{C_j}\sum_{k=1}^{C_k}p_{jk}\log \frac{p_{jk}}{p_{j.}p_{.k}}

The base of the logarithm can be the number of rows, number of columns, min(rows, columns), base-10 logarithm, natural logarithm (e), etc.

Value

A matrix of mutual information values with exametrika class. Each element (i,j) represents the mutual information between items i and j, measured in bits. Higher values indicate stronger interdependence between items.

Note

This function is implemented using a binary data compatibility wrapper and will raise an error if used with polytomous data.

Examples

# example code
# Calculate Mutual Information using sample dataset J15S500
MutualInformation(J15S500)

Omega Coefficient

Description

This function computes Tau-Congeneric Measurement, also known as McDonald's tau coefficient, for a given data set.

Usage

OmegaCoefficient(x, na = NULL, Z = NULL, w = NULL)

Arguments

Value

For a correlation/covariance matrix input, returns a single numeric value representing the omega coefficient. For a data matrix input, returns a list with three components:

OmegaCov

Omega coefficient calculated from covariance matrix

OmegaPhi

Omega coefficient calculated from phi coefficient matrix

OmegaTetrachoric

Omega coefficient calculated from tetrachoric correlation matrix

References

McDonald, R. P. (1999). Test theory: A unified treatment. Erlbaum.

internal functions for PSD of Item parameters

Description

internal functions for PSD of Item parameters

Usage

PSD_item_params(model, Lambda, quadrature, marginal_posttheta)

Arguments

Phi-Coefficient

Description

The phi coefficient is the Pearson's product moment correlation coefficient between two binary items. This function is applicable only to binary response data. The coefficient ranges from -1 to 1, where 1 indicates perfect positive correlation, -1 indicates perfect negative correlation, and 0 indicates no correlation.

Usage

PhiCoefficient(U, na = NULL, Z = NULL, w = NULL)

## Default S3 method:
PhiCoefficient(U, na = NULL, Z = NULL, w = NULL)

## S3 method for class 'binary'
PhiCoefficient(U, na = NULL, Z = NULL, w = NULL)

Arguments

Value

A matrix of phi coefficients with exametrika class. Each element (i,j) represents the phi coefficient between items i and j. The matrix is symmetric with ones on the diagonal.

Note

This function is implemented using a binary data compatibility wrapper and will raise an error if used with polytomous data.

Examples

# example code
# Calculate Phi-Coefficient using sample dataset J15S500
PhiCoefficient(J15S500)

Polychoric Correlation Matrix

Description

Polychoric Correlation Matrix

Usage

PolychoricCorrelationMatrix(U, na = NULL, Z = NULL, w = NULL)

## Default S3 method:
PolychoricCorrelationMatrix(U, na = NULL, Z = NULL, w = NULL)

## S3 method for class 'ordinal'
PolychoricCorrelationMatrix(U, na = NULL, Z = NULL, w = NULL)

Arguments

Value

A matrix of polychoric correlations with exametrika class. Each element (i,j) represents the polychoric correlation between items i and j. The matrix is symmetric with ones on the diagonal.

Examples

# example code
PolychoricCorrelationMatrix(J5S1000)

Rasch Model

Description

The one-parameter logistic model is a model with only one parameter b. This model is a 2PLM model in which a is constrained to 1. This model is also called the Rasch model.

Usage

RaschModel(b, theta)

Arguments

Value

Returns a numeric vector of probabilities between 0 and 1, representing the probability of a correct response given the ability level theta. The probability is calculated using the formula: P(\theta) = \frac{1}{1 + e^{-(\theta-b)}}

Generate Score Report for Non-Binary Test Data

Description

Calculates comprehensive descriptive statistics for a test, including measures of central tendency, variability, distribution shape, and reliability.

Usage

ScoreReport(U, na = NULL, Z = NULL, w = NULL)

Arguments

Details

This function is intended for non-binary (ordinal or rated) response data. It calculates descriptive statistics for the overall test performance. If binary data is provided, an error message will be displayed.

Value

An object of class "exametrika" and "TestStatistics" containing:

TestLength

Number of items included in the test

SampleSize

Number of examinees (rows) in the dataset

Mean

Average score across all examinees

Median

Median score

SD

Standard deviation of test scores

Variance

Variance of test scores

Skewness

Skewness of the score distribution (measure of asymmetry)

Kurtosis

Kurtosis of the score distribution (measure of tail extremity)

Min

Minimum score obtained

Max

Maximum score obtained

Range

Difference between maximum and minimum scores

Alpha

Cronbach's alpha coefficient, a measure of internal consistency reliability

Examples

# Generate score report for sample ordinal data
ScoreReport(J15S3810)

# Example with rated data
ScoreReport(J35S5000)

Structure Learning for BNM by simple GA

Description

Generating a DAG from data using a genetic algorithm.

Usage

StrLearningGA_BNM(
  U,
  Z = NULL,
  w = NULL,
  na = NULL,
  seed = 123,
  population = 20,
  Rs = 0.5,
  Rm = 0.005,
  maxParents = 2,
  maxGeneration = 100,
  successiveLimit = 5,
  crossover = 0,
  elitism = 0,
  filename = NULL,
  verbose = TRUE
)

Arguments

Details

This function generates a DAG from data using a genetic algorithm. Depending on the size of the data and the settings, the computation may take a significant amount of computational time. For details on the settings or algorithm, see Shojima(2022), section 8.5

Value

adj

Optimal adjacency matrix

testlength

Length of the test. The number of items included in the test.

TestFitIndices

Overall fit index for the test.See also TestFit

nobs

Sample size. The number of rows in the dataset.

testlength

Length of the test. The number of items included in the test.

crr

correct response ratio

TestFitIndices

Overall fit index for the test.See also TestFit

adj

Adjacency matrix

param

Learned Parameters

CCRR_table

Correct Response Rate tables

Examples

# Perform Structure Learning for Bayesian Network Model using Genetic Algorithm
# Parameters are set for balanced exploration and computational efficiency
StrLearningGA_BNM(J5S10,
  population = 20, # Size of population in each generation
  Rs = 0.5, # 50% survival rate for next generation
  Rm = 0.002, # 0.2% mutation rate for genetic diversity
  maxParents = 2, # Maximum of 2 parent nodes per item
  maxGeneration = 100, # Maximum number of evolutionary steps
  crossover = 2, # Use two-point crossover method
  elitism = 2 # Keep 2 best solutions in each generation
)

Structure Learning for BNM by PBIL

Description

Generating a DAG from data using a Population-Based Incremental Learning

Usage

StrLearningPBIL_BNM(
  U,
  Z = NULL,
  w = NULL,
  na = NULL,
  seed = 123,
  population = 20,
  Rs = 0.5,
  Rm = 0.002,
  maxParents = 2,
  maxGeneration = 100,
  successiveLimit = 5,
  elitism = 0,
  alpha = 0.05,
  estimate = 1,
  filename = NULL,
  verbose = TRUE
)

Arguments

Details

This function performs structural learning using the Population-Based Incremental Learning model(PBIL) proposed by Fukuda et al.(2014) within the genetic algorithm framework. Instead of learning the adjacency matrix itself, the 'genes of genes' that generate the adjacency matrix are updated with each generation. For more details, please refer to Fukuda(2014) and Section 8.5.2 of the text(Shojima,2022).

Value

adj

Optimal adjacency matrix

testlength

Length of the test. The number of items included in the test.

TestFitIndices

Overall fit index for the test.See also TestFit

nobs

Sample size. The number of rows in the dataset.

testlength

Length of the test. The number of items included in the test.

crr

correct response ratio

TestFitIndices

Overall fit index for the test.See also TestFit

param

Learned Parameters

CCRR_table

Correct Response Rate tables

References

Fukuda, S., Yamanaka, Y., & Yoshihiro, T. (2014). A Probability-based evolutionary algorithm with mutations to learn Bayesian networks. International Journal of Artificial Intelligence and Interactive Multimedia, 3, 7–13. DOI: 10.9781/ijimai.2014.311

Examples

# Perform Structure Learning for Bayesian Network Model using PBIL
# (Population-Based Incremental Learning)
StrLearningPBIL_BNM(J5S10,
  population = 20, # Size of population in each generation
  Rs = 0.5, # 50% survival rate for next generation
  Rm = 0.005, # 0.5% mutation rate for genetic diversity
  maxParents = 2, # Maximum of 2 parent nodes per item
  alpha = 0.05, # Learning rate for probability update
  estimate = 4 # Use rounded generational gene method
)

Structure Learning for LDLRA by PBIL algorithm

Description

Generating DAG list from data using Population-Based Incremental learning

Usage

StrLearningPBIL_LDLRA(
  U,
  Z = NULL,
  w = NULL,
  na = NULL,
  seed = 123,
  ncls = 2,
  method = "R",
  population = 20,
  Rs = 0.5,
  Rm = 0.002,
  maxParents = 2,
  maxGeneration = 100,
  successiveLimit = 5,
  elitism = 0,
  alpha = 0.05,
  estimate = 1,
  filename = NULL,
  verbose = TRUE
)

Arguments

Details

This function performs structural learning for each classes by using the Population-Based Incremental Learning model(PBIL) proposed by Fukuda et al.(2014) within the genetic algorithm framework. Instead of learning the adjacency matrix itself, the 'genes of genes' that generate the adjacency matrix are updated with each generation. For more details, please refer to Fukuda(2014) and Section 9.4.3 of the text(Shojima,2022).

Value

nobs

Sample size. The number of rows in the dataset.

testlength

Length of the test. The number of items included in the test.

crr

correct response ratio

adj_list

adjacency matrix list

g_list

graph list

referenceMatrix

Learned Parameters.A three-dimensional array of patterns where item x rank x pattern.

IRP

Marginal Item Reference Matrix

IRPIndex

IRP Indices which include Alpha, Beta, Gamma.

TRP

Test Reference Profile matrix.

LRD

latent Rank/Class Distribution

RMD

Rank/Class Membership Distribution

TestFitIndices

Overall fit index for the test.See also TestFit

Estimation_table

Estimated parameters tables.

CCRR_table

Correct Response Rate tables

Studens

Student information. It includes estimated class membership, probability of class membership, RUO, and RDO.

References

Fukuda, S., Yamanaka, Y., & Yoshihiro, T. (2014). A Probability-based evolutionary algorithm with mutations to learn Bayesian networks. International Journal of Artificial Intelligence and Interactive Multimedia, 3, 7–13. DOI: 10.9781/ijimai.2014.311

Examples

# Perform Structure Learning for LDLRA using PBIL algorithm
# This process may take considerable time due to evolutionary optimization
result.LDLRA.PBIL <- StrLearningPBIL_LDLRA(J35S515,
  seed = 123, # Set random seed for reproducibility
  ncls = 5, # Number of latent ranks
  maxGeneration = 10,
  method = "R", # Use rank model (vs. class model)
  elitism = 1, # Keep best solution in each generation
  successiveLimit = 15 # Convergence criterion
)

# Examine the learned network structure
# Plot Item Response Profiles showing item patterns across ranks
plot(result.LDLRA.PBIL, type = "IRP", nc = 4, nr = 3)

# Plot Test Response Profile showing overall response patterns
plot(result.LDLRA.PBIL, type = "TRP")

# Plot Latent Rank Distribution showing student distribution
plot(result.LDLRA.PBIL, type = "LRD")

StudentAnalysis

Description

The StudentAnalysis function returns descriptive statistics for each individual student. Specifically, it provides the number of responses, the number of correct answers, the passage rate, the standardized score, the percentile, and the stanine.

Usage

StudentAnalysis(U, na = NULL, Z = NULL, w = NULL)

Arguments

Value

Returns a data frame containing the following columns for each student:

• ID: Student identifier

• NR: Number of responses

• NRS: Number-right score (total correct answers)

• PR: Passage rate (proportion correct)

• SS: Standardized score (z-score)

• Percentile: Student's percentile rank

• Stanine: Student's stanine score (1-9)

Examples

# using sample dataset
StudentAnalysis(J15S500)

Model Fit Functions for test whole

Description

A general function that returns the model fit indices.

Usage

TestFit(U, Z, ell_A, nparam)

Arguments

Value

model_log_like

log likelihood of analysis model

bench_log_like

log likelihood of benchmark model

null_log_like

log likelihood of null model

model_Chi_sq

Chi-Square statistics for analysis model

null_Chi_sq

Chi-Square statistics for null model

model_df

degrees of freedom of analysis model

null_df

degrees of freedom of null model

NFI

Normed Fit Index. Lager values closer to 1.0 indicate a better fit.

RFI

Relative Fit Index. Lager values closer to 1.0 indicate a better fit.

IFI

Incremental Fit Index. Lager values closer to 1.0 indicate a better fit.

TLI

Tucker-Lewis Index. Lager values closer to 1.0 indicate a better fit.

CFI

Comparative Fit Index. Lager values closer to 1.0 indicate a better fit.

RMSEA

Root Mean Square Error of Approximation. Smaller values closer to 0.0 indicate a better fit.

AIC

Akaike Information Criterion. A lower value indicates a better fit.

CAIC

Consistent AIC.A lower value indicates a better fit.

BIC

Bayesian Information Criterion. A lower value indicates a better fit.

Model Fit Functions for saturated model

Description

A general function that returns the model fit indices.

Usage

TestFitSaturated(U, Z, ell_A, nparam)

Arguments

Value

model_log_like

log likelihood of analysis model

bench_log_like

log likelihood of benchmark model

null_log_like

log likelihood of null model

model_Chi_sq

Chi-Square statistics for analysis model

null_Chi_sq

Chi-Square statistics for null model

model_df

degrees of freedom of analysis model

null_df

degrees of freedom of null model

NFI

Normed Fit Index. Lager values closer to 1.0 indicate a better fit.

RFI

Relative Fit Index. Lager values closer to 1.0 indicate a better fit.

IFI

Incremental Fit Index. Lager values closer to 1.0 indicate a better fit.

TLI

Tucker-Lewis Index. Lager values closer to 1.0 indicate a better fit.

CFI

Comparative Fit Index. Lager values closer to 1.0 indicate a better fit.

RMSEA

Root Mean Square Error of Approximation. Smaller values closer to 0.0 indicate a better fit.

AIC

Akaike Information Criterion. A lower value indicates a better fit.

CAIC

Consistent AIC.A lower value indicates a better fit.

BIC

Bayesian Information Criterion. A lower value indicates a better fit.

TIF for IRT

Description

Test Information Function for 4PLM

Usage

TestInformationFunc(params, theta)

Arguments

Value

Returns a numeric vector representing the test information at each ability level theta. The test information is the sum of item information functions for all items in the test: I_{test}(\theta) = \sum_{j=1}^n I_j(\theta)

TRF for IRT

Description

Calculates the expected score across all items on a test for a given ability level (theta) using Item Response Theory. The Test Response Function (TRF) is essentially the sum of the Item Characteristic Curves (ICCs) for all items in the test.

Usage

TestResponseFunc(params, theta)

Arguments

Details

The Test Response Function computes the expected total score for an examinee with a given ability level (theta) across all items in the test. For each item, the function uses the logistic model with parameters a (discrimination), b (difficulty), c (guessing), and d (upper asymptote).

Value

A numeric vector with the same length as theta, containing the expected total score for each ability level.

Simple Test Statistics

Description

Calculates descriptive statistics for test scores, providing a comprehensive summary of central tendency, variability, and distribution shape. Different statistics are calculated based on the data type (binary, ordinal, rated, or nominal).

Usage

TestStatistics(U, na = NULL, Z = NULL, w = NULL)

## Default S3 method:
TestStatistics(U, na = NULL, Z = NULL, w = NULL)

## S3 method for class 'binary'
TestStatistics(U, na = NULL, Z = NULL, w = NULL)

## S3 method for class 'ordinal'
TestStatistics(U, na = NULL, Z = NULL, w = NULL)

Arguments

Value

The returned object depends on the data type:

For binary data, a list of class c("exametrika", "TestStatistics") containing:

TestLength

Length of the test. The number of items included in the test.

SampleSize

Sample size. The number of rows in the dataset.

Mean

Average number of correct answers.

SEofMean

Standard error of mean.

Variance

Variance of test scores.

SD

Standard Deviation of test scores.

Skewness

Skewness of score distribution (measure of asymmetry).

Kurtosis

Kurtosis of score distribution (measure of tail extremity).

Min

Minimum score.

Max

Maximum score.

Range

Range of scores (Max - Min).

Q1

First quartile. Same as the 25th percentile.

Median

Median. Same as the 50th percentile.

Q3

Third quartile. Same as the 75th percentile.

IQR

Interquartile range. Calculated by subtracting Q1 from Q3.

Stanine

Stanine score boundaries, see stanine.

For ordinal and rated data, the function calls ScoreReport and returns its result. See ScoreReport for details of the returned object.

For nominal data, an error is returned as this function does not support nominal data.

Examples

# Basic usage
stats <- TestStatistics(J15S500)
print(stats)

# Extract specific statistics
cat("Mean score:", stats$Mean, "\n")
cat("Standard deviation:", stats$SD, "\n")

# View score distribution summary
summary_stats <- data.frame(
  Min = stats$Min,
  Q1 = stats$Q1,
  Median = stats$Median,
  Mean = stats$Mean,
  Q3 = stats$Q3,
  Max = stats$Max
)
print(summary_stats)

Tetrachoric Correlation Matrix

Description

Calculates the matrix of tetrachoric correlations between all pairs of items. Tetrachoric Correlation is superior to the phi coefficient as a measure of the relation of an item pair. This function is applicable only to binary response data.

Usage

TetrachoricCorrelationMatrix(U, na = NULL, Z = NULL, w = NULL)

## Default S3 method:
TetrachoricCorrelationMatrix(U, na = NULL, Z = NULL, w = NULL)

## S3 method for class 'binary'
TetrachoricCorrelationMatrix(U, na = NULL, Z = NULL, w = NULL)

Arguments

Value

A matrix of tetrachoric correlations with exametrika class. Each element (i,j) represents the tetrachoric correlation between items i and j. The matrix is symmetric with ones on the diagonal.

Note

This function is implemented using a binary data compatibility wrapper and will raise an error if used with polytomous data.

Examples

# example code
TetrachoricCorrelationMatrix(J15S500)

Three-Parameter Logistic Model

Description

The three-parameter logistic model is a model where the lower asymptote parameter c is added to the 2PLM

Usage

ThreePLM(a, b, c, theta)

Arguments

Value

Returns a numeric vector of probabilities between c and 1, representing the probability of a correct response given the ability level theta. The probability is calculated using the formula: P(\theta) = c + \frac{1-c}{1 + e^{-a(\theta-b)}}

Two-Parameter Logistic Model

Description

The two-parameter logistic model is a classic model that defines the probability of a student with ability theta successfully answering item j, using both a slope parameter and a location parameter.

Usage

TwoPLM(a, b, theta)

Arguments

Value

Returns a numeric vector of probabilities between 0 and 1, representing the probability of a correct response given the ability level theta. The probability is calculated using the formula: P(\theta) = \frac{1}{1 + e^{-a(\theta-b)}}

Prior distribution function with guessing parameter

Description

Prior distribution function with guessing parameter

Usage

asymprior(c, alp, bet)

Arguments

calc Fit Indices

Description

A general function that returns the model fit indices.

Usage

calcFitIndices(chi_A, chi_B, df_A, df_B, nobs)

Arguments

Value

NFI

Normed Fit Index. Lager values closer to 1.0 indicate a better fit.

RFI

Relative Fit Index. Lager values closer to 1.0 indicate a better fit.

IFI

Incremental Fit Index. Lager values closer to 1.0 indicate a better fit.

TLI

Tucker-Lewis Index. Lager values closer to 1.0 indicate a better fit.

CFI

Comparative Fit Index. Lager values closer to 1.0 indicate a better fit.

RMSEA

Root Mean Square Error of Approximation. Smaller values closer to 0.0 indicate a better fit.

AIC

Akaike Information Criterion. A lower value indicates a better fit.

CAIC

Consistent AIC.A lower value indicates a better fit.

BIC

Bayesian Information Criterion. A lower value indicates a better fit.

Correct Response Rate

Description

The correct response rate (CRR) is one of the most basic and important statistics for item analysis. This is an index of item difficulty and a measure of how many students out of those who tried an item correctly responded to it. This function is applicable only to binary response data.

The CRR for each item is calculated as:

p_j = \frac{\sum_{i=1}^n z_{ij}u_{ij}}{\sum_{i=1}^n z_{ij}}

where z_{ij} is the missing indicator and u_{ij} is the response.

Usage

crr(U, na = NULL, Z = NULL, w = NULL)

## Default S3 method:
crr(U, na = NULL, Z = NULL, w = NULL)

## S3 method for class 'binary'
crr(U, na = NULL, Z = NULL, w = NULL)

Arguments

Value

A numeric vector of weighted correct response rates for each item. Values range from 0 to 1, where higher values indicate easier items (more students answered correctly).

Note

This function is implemented using a binary data compatibility wrapper and will raise an error if used with polytomous data.

Examples

# Simple binary data
U <- matrix(c(1, 0, 1, 1, 0, 1), ncol = 2)
crr(U)

# using sample datasaet
crr(J15S500)

dataFormat

Description

This function serves the role of formatting the data prior to the analysis.

Usage

dataFormat(
  data,
  na = NULL,
  id = 1,
  Z = NULL,
  w = NULL,
  response.type = NULL,
  CA = NULL
)

Arguments

Value

U

For binary response data. A matrix with rows representing the sample size and columns representing the number of items, where elements are either 0 or 1. u_{ij}=1 indicates that student i correctly answered item j, while u_{ij}=0 means that student i answered item j incorrectly.

Q

For polytomous response data. A matrix with rows representing the sample size and columns representing the number of items, where elements are non-negative integers. When input data is in factor format, the factor levels are converted to consecutive integers starting from 1.

ID

The ID label given by the designated column or function.

ItemLabel

The item names given by the provided column names or function.

Z

Missing indicator matrix. z_{ij}=1 indicates that item j is presented to Student i, while z_{ij}=0 indicates item j is NOT presented to Student i. If the data contains NA values, -1 is assigned.

w

Item weight vector

response.type

Character string indicating the type of response data: "binary", "ordinal", "rated", or "nominal"

CategoryLabel

List containing the original factor labels when polytomous responses are provided as factors. NULL if no factor data is present.

categories

Numeric vector containing the number of response categories for each item.

CA

For rated polytomous data, a numeric vector of correct answers. NULL for other types.

Generate category labels for response data

Description

Generate category labels for response data

Usage

generate_category_labels(data_column, item_name)

Arguments

Details

If the input is a factor, returns its levels. Otherwise, generates labels in the format "Item name-Category-N"

Value

A character vector of category labels

generate start values for optimize

Description

generate start values for optimize

Usage

generate_start_values(tmp)

Arguments

cumulative probability of GRM

Description

cumulative probability of GRM

Usage

grm_cumprob(theta, a, b)

Arguments

Item Information Function for GRM

Description

Calculates the value of the Item Information Function for the Graded Response Model.

Usage

grm_iif(theta, a, b)

Arguments

Value

Value of the Item Information Function

Examples

## Not run: 
# Example for an item with 3 categories
a <- 1.5
b <- c(-1.0, 1.0)
thetas <- seq(-3, 3, by = 0.1)
info <- sapply(thetas, function(t) grm_iif(t, a, b))
plot(thetas, info, type = "l", xlab = "Theta", ylab = "Information")

## End(Not run)

Probability function for GRM

Description

Calculates the probability of selecting each category given a latent trait value and item parameters.

Usage

grm_prob(theta, a, b)

Arguments

Value

Vector of category selection probabilities

Examples

## Not run: 
# Example for an item with 3 categories
a <- 1.5
b <- c(-1.0, 1.0)
theta <- 0
grm_prob(theta, a, b)

## End(Not run)

log_lik function for grm

Description

log_lik function for grm

Usage

log_lik_grm(target, dat, verbose)

Arguments

Long Format Data Conversion

Description

A function to reshape long data into a dataset suitable for exametrika.

Usage

longdataFormat(
  data,
  na = NULL,
  Sid = NULL,
  Qid = NULL,
  Resp = NULL,
  w = NULL,
  response.type = NULL,
  CA = NULL
)

Arguments

Value

U

For binary response data. A matrix with rows representing the sample size and columns representing the number of items, where elements are either 0 or 1. u_{ij}=1 indicates that student i correctly answered item j, while u_{ij}=0 means that student i answered item j incorrectly.

Q

For polytomous response data. A matrix with rows representing the sample size and columns representing the number of items, where elements are non-negative integers. When input data is in factor format, the factor levels are converted to consecutive integers starting from 1.

ID

The ID label given by the designated column or function.

ItemLabel

The item names given by the provided column names or function.

Z

Missing indicator matrix. z_{ij}=1 indicates that item j is presented to Student i, while z_{ij}=0 indicates item j is NOT presented to Student i.

w

Item weight vector

response.type

Character string indicating the type of response data: "binary", "ordinal", "rated", or "nominal"

CategoryLabel

List containing the original factor labels when polytomous responses are provided as factors. NULL if no factor data is present.

categories

Numeric vector containing the number of response categories for each item.

CA

For rated polytomous data, a numeric vector of correct answers. NULL for other types.

Utility function for searching DAG

Description

Function to limit the number of parent nodes

Usage

maxParents_penalty(vec, testlength, maxParents)

Arguments

Details

When generating an adjacency matrix using GA, the number of edges coming from a single node should be limited to 2 or 3. This is because if there are too many edges, it becomes difficult to interpret in practical applications. This function works to adjust the sampling of the randomly generated adjacency matrix so that the column sum of the upper triangular elements fits within the set limit.

Number Right Score

Description

The Number-Right Score (NRS) function calculates the weighted sum of correct responses for each examinee. This function is applicable only to binary response data.

For each examinee, the score is computed as:

NRS_i = \sum_{j=1}^J z_{ij}u_{ij}w_j

where:

• z_{ij} is the missing response indicator (0/1)

• u_{ij} is the response (0/1)

• w_j is the item weight

Usage

nrs(U, na = NULL, Z = NULL, w = NULL)

## Default S3 method:
nrs(U, na = NULL, Z = NULL, w = NULL)

## S3 method for class 'binary'
nrs(U, na = NULL, Z = NULL, w = NULL)

Arguments

Value

A numeric vector containing the Number-Right Score for each examinee. The score represents the weighted sum of correct answers, where:

• Maximum score is the sum of all item weights

• Minimum score is 0

• Missing responses do not contribute to the score

Note

This function is implemented using a binary data compatibility wrapper and will raise an error if used with polytomous data.

Examples

# using sample dataset
nrs(J15S500)

Log-likelihood function used in the Maximization Step (M-Step).

Description

Log-likelihood function used in the Maximization Step (M-Step).

Usage

objective_function_IRT(lambda, model, qjtrue, qjfalse, quadrature)

Arguments

parameter transformation params_to_target

Description

parameter transformation params_to_target

Usage

params_to_target_jac(a_vec, b_list)

Arguments

Passage Rate of Student

Description

The Passage Rate for each student is calculated as their Number-Right Score (NRS) divided by the number of items presented to them. This function is applicable only to binary response data.

The passage rate is calculated as:

P_i = \frac{\sum_{j=1}^J z_{ij}u_{ij}w_j}{\sum_{j=1}^J z_{ij}}

where:

• z_{ij} is the missing response indicator (0/1)

• u_{ij} is the response (0/1)

• w_j is the item weight

Usage

passage(U, na = NULL, Z = NULL, w = NULL)

## Default S3 method:
passage(U, na = NULL, Z = NULL, w = NULL)

## S3 method for class 'binary'
passage(U, na = NULL, Z = NULL, w = NULL)

Arguments

Value

A numeric vector containing the passage rate for each student. Values range from 0 to 1 (or maximum weight) where:

• 1: Perfect score on all attempted items

• 0: No correct answers

• NA: No items attempted

Note

This function is implemented using a binary data compatibility wrapper and will raise an error if used with polytomous data.

The passage rate accounts for missing responses by only considering items that were actually presented to each student. This provides a fair comparison between students who attempted different numbers of items.

Examples

# using sample dataset
passage(J15S500)

Student Percentile Ranks

Description

The percentile function calculates each student's relative standing in the group, expressed as a percentile rank (1-100). This function is applicable only to binary response data.

The percentile rank indicates the percentage of scores in the distribution that fall below a given score. For example, a percentile rank of 75 means the student performed better than 75% of the group.

Usage

percentile(U, na = NULL, Z = NULL, w = NULL)

## Default S3 method:
percentile(U, na = NULL, Z = NULL, w = NULL)

## S3 method for class 'binary'
percentile(U, na = NULL, Z = NULL, w = NULL)

Arguments

Value

A numeric vector of percentile ranks (1-100) for each student, where:

• 100: Highest performing student(s)

• 50: Median performance

• 1: Lowest performing student(s)

Note

This function is implemented using a binary data compatibility wrapper and will raise an error if used with polytomous data.

Percentile ranks are calculated using the empirical cumulative distribution function of standardized scores. Tied scores receive the same percentile rank. The values are rounded up to the nearest integer to provide ranks from 1 to 100.

Examples

# using sample dataset
percentile(J5S10)

Plot Method for Objects of Class "exametrika"

Description

Creates visualizations for objects with class "exametrika". The calculation results of the exametrika package have an exametrika class attribute, along with the specific analysis model class (IRT, GRM, LCA, LRA, Biclustering, IRM, LDLRA, LDB, BINET). Each model has its own compatible plot types, accessible by specifying the 'type' parameter.

Usage

## S3 method for class 'exametrika'
plot(
  x,
  type = c("IRF", "TRF", "IIF", "TIF", "IIC", "ICC", "TIC", "IRP", "TRP", "LCD", "CMP",
    "FRP", "RMP", "LRD", "Array", "CRV", "RRV", "FieldPIRP", "LDPSR", "ScoreFreq",
    "ScoreRank", "ICRP", "ICBR"),
  items = NULL,
  students = NULL,
  nc = 1,
  nr = 1,
  overlay = FALSE,
  ...
)

Arguments

Details

Each model class supports specific plot types:

IRT

Supports "IRF"/"ICC", "TRF", "IIF"/"IIC", "TIF"/"TIC"

GRM

Supports "IRF"/"ICC", "IIF"/"IIC", "TIF"/"TIC"

LCA

Supports "IRP", "FRP", "TRP", "LCD", "CMP"

LRA

Supports "IRP", "FRP", "TRP", "LRD", "RMP"

LRAordinal

Supports "ScoreFreq", "ScoreRank", "ICRP", "ICBR", "RMP"

LRArated

Supports "ScoreFreq", "ScoreRank", "ICRP", "RMP"

Biclustering

Supports "FRP", "TRP", "LCD", "LRD", "CMP", "RMP", "CRV", "RRV", "Array"

IRM

Supports "FRP", "TRP", "Array"

LDLRA

Supports "IRP", "TRP", "LRD", "RMP"

LDB

Supports "FRP", "TRP", "LRD", "RMP", "Array", "FieldPIRP"

BINET

Supports "FRP", "TRP", "LRD", "RMP", "Array", "LDPSR"

Value

Produces visualizations based on the model class and specified type:

IRT models

IRF (Item Response Function), TRF (Test Response Function), IIF (Item Information Function), TIF (Test Information Function)

LCA/LRA models

IRP (Item Reference Profile), TRP (Test Reference Profile), LCD/LRD (Latent Class/Rank Distribution), CMP/RMP (Class/Rank Membership Profile)

Biclustering/IRM models

Array plots showing clustering patterns, FRP, TRP, etc.

LDLRA/LDB/BINET models

Network and profile plots specific to each model

Examples

## Not run: 
# IRT model example
irt_result <- exametrika::IRT(U)
plot(irt_result, type = "IRF", items = 1:5)
plot(irt_result, type = "TIF")

# LCA model example
lca_result <- exametrika::LCA(U)
plot(lca_result, type = "IRP")
plot(lca_result, type = "LCD")

## End(Not run)

Polychoric Correlation

Description

Calculate the polychoric correlation coefficient between two polytomous (categorical ordinal) variables. Polychoric correlation estimates the correlation between two theorized normally distributed continuous latent variables from two observed ordinal variables.

Usage

polychoric(x, y)

Arguments

Details

This function handles missing values (coded as -1 or NA) using pairwise deletion. The estimation uses maximum likelihood approach with Brent's method for optimization.

Value

The polychoric correlation coefficient between x and y

Examples

# Example with simulated data
set.seed(123)
x <- sample(1:5, 100, replace = TRUE)
y <- sample(1:4, 100, replace = TRUE)
polychoric(x, y)

Calculate Polychoric Correlation Likelihood

Description

Calculates the negative log-likelihood for estimating polychoric correlation from a contingency table of two ordinal variables.

Usage

polychoric_likelihood(rho, mat)

Arguments

Details

The function estimates thresholds from the marginal distributions and calculates the expected probabilities based on a bivariate normal distribution. It then computes the log-likelihood by comparing observed and expected frequencies.

Value

The negative log-likelihood value for the given correlation coefficient

Polyserial Correlation

Description

Calculates the polyserial correlation coefficient between a continuous variable and an ordinal variable.

Usage

polyserial(x, y)

Arguments

Details

This function implements Olsson et al.'s ad hoc method for estimating the polyserial correlation coefficient. The method assumes that the continuous variable follows a normal distribution and that the ordinal variable is derived from an underlying continuous normal variable through thresholds.

Value

A numeric value representing the estimated polyserial correlation coefficient.

References

U.Olsson, F.Drasgow, and N.Dorans (1982). The polyserial correlation coefficient. Psychometrika, 47,337-347.

Examples

n <- 300
x <- rnorm(n)
y <- sample(1:5, size = n, replace = TRUE)
polyserial(x, y)

Print Method for Exametrika Objects

Description

S3 method for printing objects of class "exametrika". This function formats and displays appropriate summary information based on the specific subclass of the exametrika object. Different types of analysis results (IRT, LCA, network models, etc.) are presented with customized formatting to highlight the most relevant information.

Usage

## S3 method for class 'exametrika'
print(x, digits = 3, ...)

Arguments

Details

The function identifies the specific subclass of the exametrika object and tailors the output accordingly. For most analysis types, the function displays:

• Basic model description and parameters

• Estimation results (e.g., item parameters, latent class profiles)

• Model fit statistics and diagnostics

• Visual representations where appropriate (e.g., graphs for network models, scree plots for dimensionality analysis)

When printing network-based models (LDLRA, LDB, BINET), this function visualizes the network structure using graphs, which can help in interpreting complex relationships between items or latent variables.

Value

Prints a formatted summary of the exametrika object to the console, with content varying by object subclass:

TestStatistics

Basic descriptive statistics of the test

Dimensionality

Eigenvalue analysis results with scree plot

ItemStatistics

Item-level statistics and psychometric properties

QitemStatistics

Item statistics for polytomous items

exametrikaData

Data structure details including response patterns and weights

IIAnalysis

Item-item relationship measures (tetrachoric correlations, etc.)

CTT

Classical Test Theory reliability measures

IRT/GRM

Item parameters, ability estimates, and fit indices

LCA/LRA

Class/Rank profiles, distribution information, and model fit statistics

Biclustering/IRM

Cluster profiles, field distributions, and model diagnostics

LDLRA/LDB/BINET

Network visualizations, parameter estimates, and conditional probabilities

Examples

# Print IRT analysis results with 4 decimal places
result <- IRT(J15S500)
print(result, digits = 4)

# Print Latent Class Analysis results
result_lca <- LCA(J15S500, ncls = 3)
print(result_lca)

bivariate normal CDF

Description

Calculates the cumulative distribution function (CDF) of a bivariate normal distribution. This function computes P(X <= a, Y <= b) where X and Y follow a bivariate normal distribution with correlation coefficient rho.

Usage

qBiNormal(a, b, rho)

Arguments

Details

The implementation uses numerical integration with Gauss-Legendre quadrature for accurate computation. Special cases for infinite bounds are handled separately.

Value

The probability P(X <= a, Y <= b), a value between 0 and 1.

Generate Error Message for Invalid Response Type

Description

Internal function to generate standardized error messages when a function is called with an incompatible response type.

Usage

response_type_error(response_type, fun_name)

Arguments

Value

Never returns; always stops with an error message

score function for grm

Description

score function for grm

Usage

score_function_with_Jacobian(target, dat)

Arguments

Prior distribution function with respect to the slope.

Description

Prior distribution function with respect to the slope.

Usage

slopeprior(a, m, s, const = 1e-15)

Arguments

softmax function

Description

to avoid overflow

Usage

softmax(x)

Arguments

Standardized Score

Description

The standardized score (z-score) indicates how far a student's performance deviates from the mean in units of standard deviation. This function is applicable only to binary response data.

The score is calculated by standardizing the passage rates:

Z_i = \frac{r_i - \bar{r}}{\sigma_r}

where:

• r_i is student i's passage rate

• \bar{r} is the mean passage rate

• \sigma_r is the standard deviation of passage rates

Usage

sscore(U, na = NULL, Z = NULL, w = NULL)

## Default S3 method:
sscore(U, na = NULL, Z = NULL, w = NULL)

## S3 method for class 'binary'
sscore(U, na = NULL, Z = NULL, w = NULL)

Arguments

Value

A numeric vector of standardized scores for each student. The scores follow a standard normal distribution with:

• Mean = 0

• Standard deviation = 1

• Approximately 68% of scores between -1 and 1

• Approximately 95% of scores between -2 and 2

• Approximately 99% of scores between -3 and 3

Note

This function is implemented using a binary data compatibility wrapper and will raise an error if used with polytomous data.

The standardization allows for comparing student performance across different tests or groups. A positive score indicates above-average performance, while a negative score indicates below-average performance.

Examples

# using sample dataset
sscore(J5S10)

Stanine Scores

Description

The Stanine (Standard Nine) scoring system divides students into nine groups based on a normalized distribution. This function is applicable only to binary response data.

These groups correspond to the following percentile ranges:

• Stanine 1: lowest 4% (percentiles 1-4)

• Stanine 2: next 7% (percentiles 5-11)

• Stanine 3: next 12% (percentiles 12-23)

• Stanine 4: next 17% (percentiles 24-40)

• Stanine 5: middle 20% (percentiles 41-60)

• Stanine 6: next 17% (percentiles 61-77)

• Stanine 7: next 12% (percentiles 78-89)

• Stanine 8: next 7% (percentiles 90-96)

• Stanine 9: highest 4% (percentiles 97-100)

Usage

stanine(U, na = NULL, Z = NULL, w = NULL)

## Default S3 method:
stanine(U, na = NULL, Z = NULL, w = NULL)

## S3 method for class 'binary'
stanine(U, na = NULL, Z = NULL, w = NULL)

Arguments

Value

A list containing two elements:

stanine

The score boundaries for each stanine level

stanineScore

The stanine score (1-9) for each student

Note

This function is implemented using a binary data compatibility wrapper and will raise an error if used with polytomous data.

Stanine scores provide a normalized scale with:

• Mean = 5

• Standard deviation = 2

• Scores range from 1 to 9

• Score of 5 represents average performance

References

Angoff, W. H. (1984). Scales, norms, and equivalent scores. Educational Testing Service. (Reprint of chapter in R. L. Thorndike (Ed.) (1971) Educational Measurement (2nd Ed.). American Council on Education.

Examples

result <- stanine(J15S500)
# View score boundaries
result$stanine
# View individual scores
result$stanineScore

parameter transformation target_to_params

Description

parameter transformation target_to_params

Usage

target_to_params_jac(target, nitems, ncat)

Arguments

Tetrachoric Correlation

Description

Tetrachoric Correlation is superior to the phi coefficient as a measure of the relation of an item pair. See Divgi, 1979; Olsson, 1979;Harris, 1988.

Usage

tetrachoric(x, y)

Arguments

Value

Returns a single numeric value of class "exametrika" representing the tetrachoric correlation coefficient between the two binary variables. The value ranges from -1 to 1, where:

• 1 indicates perfect positive correlation

• -1 indicates perfect negative correlation

• 0 indicates no correlation

References

Divgi, D. R. (1979). Calculation of the tetrachoric correlation coefficient. Psychometrika, 44, 169–172.

Olsson, U. (1979). Maximum likelihood estimation of the polychoric correlation coefficient. Psychometrika,44, 443–460.

Harris, B. (1988). Tetrachoric correlation coefficient. In L. Kotz, & N. L. Johnson (Eds.), Encyclopedia of statistical sciences (Vol. 9, pp. 223–225). Wiley.