Type:	Package
Title:	Utilities for Streamlined Data Import, Imputation and Modelling
Version:	1.1.5
Description:	Provides functions streamlining the data analysis workflow: Outsourcing data import, renaming and type casting to a *.csv. Manipulating imputed datasets and fitting models on them. Summarizing models.
Depends:	R (≥ 4.0.0)
Imports:	assertthat, dplyr, mice, Hmisc, survival, stats, purrr, MASS, sae
License:	GPL (≥ 3)
Encoding:	UTF-8
RoxygenNote:	7.2.3
Suggests:	testthat (≥ 3.0.0), rms
Config/testthat/edition:	3
URL:	https://CRAN.R-project.org/package=basecamb, https://github.com/codeblue-team/basecamb
BugReports:	https://github.com/codeblue-team/basecamb/issues
NeedsCompilation:	no
Packaged:	2024-04-21 17:58:45 UTC; Peter
Author:	J. Peter Marquardt [aut, cre], Till D. Best [aut]
Maintainer:	J. Peter Marquardt <peter@kmarquardt.de>
Repository:	CRAN
Date/Publication:	2024-04-22 19:10:07 UTC

Locate NA values introduced during apply_data_dictionary()

Description

Finds and locates NA values that were introduced by calling apply_data_dictionary() on a dataframe using a data_dictionary.

Usage

.find_NA_coercions(data_raw, data, data_dictionary)

Arguments

Value

returns a dataframe with the location of introduced NA's. If no NA's were introduced an empty dataframe is returned.

Author(s)

Till D. Best, J. Peter Marquardt

Parse a string to create a named list

Description

Create a named list from a standardised string of the following format:

• key-value pairs are separated from other key-value-pairs by a comma (",")

• key and value of the same pair are separated by an equal sign ("=")

• quotations around individual keys and values are recommended for clarity, but do not affect functionality.

• all values will be coerced to type character, with the exception of "NA", "TRUE" and "FALSE"

Usage

.parse_string_to_named_vector(str)

Arguments

Value

named vector

Author(s)

J. Peter Marquardt

Scaling a variable

Description

A helper function to scale a variable in a dataframe. Divides 'variable' by 'scaling_denominator'.

Usage

.scale_variable(data, variable, scaling_denominator)

Arguments

Value

the input dataframe with the newly scaled 'variable'

Clean column names, types and levels

Description

Use a data dictionary data.frame to apply the following tidying steps to your data.frame:

• Remove superfluous columns

• Rename columns

• Ensure/coerce correct data type for each column

• Assign factorial levels, including renaming and grouping

Usage

apply_data_dictionary(
  data,
  data_dictionary,
  na_action_default = "keep_NA",
  print_coerced_NA = TRUE
)

Arguments

Value

clean data.frame

Author(s)

J. Peter Marquardt

Apply function to dataframes in a mice object

Description

Wrapper function to apply a function on each dataframe in an imputed dataset created with mice::mice().

Usage

apply_function_to_imputed_data(mice_data, fun, ...)

Arguments

Value

a mids object with transformed data.

Author(s)

J. Peter Marquardt

Assign custom values for key levels in factorial columns

Description

Use a named vector of keys (current value) and values for factorial columns to assign meaningful levels and/or group levels

Usage

assign_factorial_levels(
  data,
  factor_keys_values,
  na_action_default = "keep_NA"
)

Arguments

Value

data frame with new levels

Author(s)

J. Peter Marquardt

Examples

data <- data.frame(col1 = as.factor(rep(c('1', '2', '4'), 5)))
keys_1 <- list('col1' = c('1' = 'One', '2' = 'Two', '4' = 'Four'))
data_1 <- assign_factorial_levels(data, keys_1)
keys_2 <- list('col1' = c('1' = 'One', 'default' = 'Not_One'))
data_2 <- assign_factorial_levels(data, keys_2)

Assign tidy types and names to a data.frame

Description

Verbosely assign tidy name and data type for each column of a data.frame and get rid of superfluous columns. Uses a .csv file for assignments to encourage a data dictionary based workflow. CAVE! Requires 'Date' type columns to already be read in as Date.

Usage

assign_types_names(data, meta_data)

Arguments

Value

clean data.frame

Author(s)

J. Peter Marquardt

Build formula for statistical models

Description

Build formula used in statistical models from vectors of strings with the option to specify an environment.

Usage

build_model_formula(
  outcome,
  predictors,
  censor_event = NULL,
  env = parent.frame()
)

Arguments

Value

formula for use in statistical models

Author(s)

J. Peter Marquardt

Examples

build_model_formula("outcome", c("pred_1", "pred_2"))
build_model_formula("outcome", c("pred_1", "pred_2"), censor_event = "cens_event")

Test cox proportional odds assumption on models using multiple imputation.

Description

Constructs a model and conducts a cox.zph test for each imputation of the data set.

Usage

cox.zph.mids(
  model,
  imputations,
  p_level = 0.05,
  global_only = TRUE,
  return_raw = FALSE,
  p_only = TRUE,
  verbose = TRUE
)

Arguments

Value

depending on specified options, this function can return

• default: A vector of global p-values

• global_only = FALSE: a data.frame with p-values for all variables plus the global

• return_raw = TRUE: list of cox.zph objects

Author(s)

J. Peter Marquardt

Examples

data <- data.frame(time = 101:200, status = rep(c(0,1), 50), pred = rep(c(1:9, NA), 10))
imputed_data <- mice::mice(data)
cox_mod <- Hmisc::fit.mult.impute(survival::Surv(time, status) ~ pred,
fitter = rms::cph, xtrans = imputed_data)
cox.zph.mids(cox_mod, imputed_data)

Deconstruct formula

Description

Deconstruct a formula object into strings of its components. Predictors are split by '+', so interaction terms will be returned as a single string.

Usage

deconstruct_formula(formula)

Arguments

Value

a named list with fields:

• outcome (character)

• predictors (vector of characters)

• censor_event (character) (optional) censor event, only for formulas including a Surv() object

Author(s)

J. Peter Marquardt

Examples

deconstruct_formula(stats::as.formula("outcome ~ predictor1 + predictor2 + predictor3"))
deconstruct_formula(stats::as.formula("Surv(outcome, censor_event) ~ predictor"))

Filter dataframe for nth entry

Description

Filter a dataframe for the nth entry of each subject in it. A typical use cases would be to filter a dataset for the first or last measurement of a subject.#'

Usage

filter_nth_entry(data, ID_column, entry_column, n = 1, reverse_order = FALSE)

Arguments

Value

data.frame with <= 1 entry per subject

Author(s)

J. Peter Marquardt

Examples

data <- data.frame(list(ID = rep(1:5, 3), encounter = rep(1:3, each=5), value = rep(4:6, each=5)))
filter_nth_entry(data, 'ID', 'encounter')
filter_nth_entry(data, 'ID', 'encounter', n = 2)
filter_nth_entry(data, 'ID', 'encounter', reverse_order = TRUE)

Fit a model on multiply imputed data using only observations with non-missing outcome(s)

Description

This function is a wrapper for fitting models with Hmisc::fit.mult.impute() on a multiply imputed dataset generated with mice::mice(). Cases with a missing outcome in the original dataset are removed from the mids object by using the "subset" argument in Hmisc::fit.mult.impute().

Usage

fit_mult_impute_obs_outcome(mids, formula, fitter, ...)

Arguments

Value

mod a fit.mult.impute object.

Author(s)

Till D. Best

Examples

# create an imputed dataset
imputed_data <- mice::mice(airquality)

fit_mult_impute_obs_outcome(mids = imputed_data, formula = Ozone ~ Solar.R + Wind, fitter = glm)

Summarise a logistic regression model on the odds ratio scale

Description

This function summarises regression models that return data on the log-odds scale and returns a dataframe with estimates, and confidence intervals as odds ratios. P value are also provided. Additionally, intercepts can be removed from the summary. This comes in handy when ordinal logistic regression models are fit. Ordinal regression models (such as proportional odds models) usually result in many intercepts that are not really of interest. This function is also compatible with models obtained from multiply imputed datasets, for example models fitted with Hmisc::fit.mult.impute().

Usage

or_model_summary(
  model,
  conf_int = 1.96,
  print_intercept = FALSE,
  round_est = 3,
  round_p = 4
)

Arguments

Details

CAVE! The function does not check whether your estimates are on the log-odds scale. It will do the transformation no matter what!

Value

a dataframe with the adjusted odds ratio, confidence intervals and p-values.

Author(s)

Till D. Best

Examples

# fit a logistic model
mod <- glm(formula = am ~ mpg + cyl, data = mtcars, family = binomial())

or_model_summary(model = mod)

Parse values in date columns as Dates

Description

Parse date columns in a data.frame as Date. Use a named list to specify each date column (key) and the format (value) it is coded in.

Usage

parse_date_columns(data, date_formats)

Arguments

Value

data.frame with date columns in Date type

Author(s)

J. Peter Marquardt

Examples

data <- data.frame(date = rep('01/23/4567', 5))
data <- parse_date_columns(data, list(date = '%m/%d/%Y'))

Stratify a numeric vector into quantile groups

Description

Transforms a numeric vector into quantile groups. For each input value, the output value corresponds to the quantile that value is in. When grouping into n quantiles, the lowest 1/n of values are assigned 1, the highest 1/n are assigned n.

Usage

quantile_group(data, n, na.rm = TRUE)

Arguments

Details

Tied values will be assigned to the lower quantile group rather than etsimating a distribution. In extreme cases this can mean one or more quantile groups are not represented.

If uneven group sizes cannot be avoided, values will be assigned the higher quantile group.

Value

vector of length length(data) with the quantile groups

Author(s)

J. Peter Marquardt

Examples

quantile_group(10:1, 3)
quantile_group(c(rep(1,3), 10:1, NA), 5)

Remove duplicate rows from data.frame

Description

Removes rows that are duplicates of another row in all columns except exclude_columns

Usage

remove_duplicates(
  data,
  exclude_columns = NULL,
  ID_column = NULL,
  quiet = FALSE
)

Arguments

Details

Wraps unique()

Value

vector of row indices with non-unique data

Author(s)

J. Peter Marquardt

Examples

data <- data.frame(Study_ID = c("A", "B", "C"), ID = c(123, 456, 123), num_cars = c(10, 2, 10))
remove_duplicates(data, exclude_columns = "Study_ID")
remove_duplicates(data, exclude_columns = "Study_ID", ID_column = "ID")

Remove missing cases from a mids object

Description

Deprecated, use apply_function_to_imputed_data instead.

Usage

remove_missing_from_mids(mids, var)

Arguments

Details

Remove_missing_from_mids is used to filter a mids object for missing cases in the original dataset in the variable var. This is useful for situations where you want to use as many observations as possible for imputation but only fit your model on a subset of these. Or, if you want to create one large imputed datset from which multiple analyses with multiple outcomes are derived.

Value

a mids object filtered for observed cases of var.

Author(s)

Till D. Best

See Also

apply_function_to_imputed_data

Scale continuous predictors

Description

This function linearly scales variables in data objects according to a data dictionary. The data dictionary has at least two columns, "variable" and "scaling_denominator". "Variable" is divided by "scaling_denominator".

Usage

scale_continuous_predictors(data, scaling_dictionary)

Arguments

Value

The data with the newly scaled 'variables'.

Author(s)

Till D. Best

Identify duplicate values in a vector representing a set

Description

Identify duplicate values in a vector representing a set

Usage

setduplicates(vect)

Arguments

Value

a vector of duplicate elements

Author(s)

J. Peter Marquardt

See Also

setops

Examples

setduplicates(c(1,2,2,3))

Box-Cox transformation for stratified data

Description

Create Box-Cox transformation using different optimal lambda values for each stratum

Usage

stratified_boxcox(
  data,
  value_col,
  strat_cols,
  plot = FALSE,
  return = "values",
  buffer = 0,
  inverse = FALSE,
  lambdas = NULL
)

Arguments

Value

if "values", vector of transformed values, if "lambdas" nested named list of used lambdas. The buffer will be equal for all strata

Author(s)

J. Peter Marquardt

Examples

data <- data.frame("value" = c(1:50, rnorm(50, 100, 10)),
                   "strat_var" = rep(c(1,2), each = 50),
                   "strat_var2" = rep(c(1, 2), 50))
lambdas <- stratified_boxcox(data = data, value_col = "value",
                             strat_cols = c("strat_var", "strat_var2"),
                             return = "lambdas")
data$value_boxed <- stratified_boxcox(data = data, value_col = "value",
                                      strat_cols = c("strat_var", "strat_var2"),
                                      return = "values")
data$value_unboxed <- stratified_boxcox(data = data, value_col = "value_boxed",
                                        strat_cols = c("strat_var", "strat_var2"),
                                        inverse = TRUE, lambdas = lambdas)