Clarke Test

Description

‘clarke_test' returns results from Kevin Clarke’s distribution-free test of non-nested models.

Usage

clarke_test(model1, model2, level=0.05, digits=2)

Arguments

Details

'clarke_test' is a more modularized version of the [clarke()] function from the [games] package. The main innovation is that the 'nonnest' function calls a generic 'indivLogLiks' function, so additional methods can be easily written for different classes of models. The function currently supports binomial, poisson and negative binomial GLMs, ordinal models estimated with either polr from the MASS package or clm from the ordinal package and multinomial models estimated with either multinom from the nnet package. Users can also write new methods for both indivLogLiks and nparams that would get called by the generic function.

Value

Typical use will be to run the function interactively and examine the printed output. The functions return an object of class nonnest.test, which is a list containing:

stat

The test statistic

level

Significance level for the test

digits

Number of digits to print

loglik1

Vector of observationwise log-likelihoods for model1

loglik2

Vector of observationwise log-likelihoods for model2

nparams

Integer vector containing the number of parameters fitted in model1 and model2 respectively

nobs

Number of observations of the dependent variable being modeled

An object of class nonnest.test with the following values:

stat

The number of times model 1 is better than model 2

test

Will always be "clarke".

level

The chosen confidence level for the test

digits

The number of digits to print

loglik1

Individual log-likelihoods for model 1

loglik2

Individual log-likelihoods for model 2

nparams

A vector giving the number of parameters in models 1 and 2, respectively

nobs

Number of observations in the model

Author(s)

Brenton Kenkel (brenton.kenkel@gmail.com) modified by Dave Armstrong (dave@quantoid.net)

References

Kevin Clarke. 2007. "A Simple Distribution-Free Test for Nonnested Hypotheses." Political Analysis 15(3): 347–363.

Examples

data(conflictData)
## Linear Model
lm1 <- lm(riots ~ log(rgdpna_pc) + log(pop*1000) +
    polity2, data=conflictData)
lm2 <- lm(riots ~ rgdpna_pc + pop +
    polity2, data=conflictData)
clarke_test(lm1, lm2)

## Binomial GLM
glm1 <- glm(conflict_binary ~ log(rgdpna_pc) +
          log(pop*1000) + polity2, data=conflictData,
          family=binomial)
glm2 <- glm(conflict_binary ~ rgdpna_pc + pop +
          polity2, data=conflictData,
          family=binomial)
clarke_test(glm1, glm2)

## Poisson GLM
glm1a <- glm(riots ~ log(rgdpna_pc) +
              log(pop*1000) + polity2,
             data=conflictData,
             family=poisson)
glm2a <- glm(riots ~ rgdpna_pc + pop +
              polity2, data=conflictData,
            family=poisson)
clarke_test(glm1a, glm2a)

## Negative Binomial GLM
library(MASS)
glm1b <- glm.nb(riots ~ log(rgdpna_pc) +
               log(pop*1000) + polity2,
               data=conflictData)
glm2b <- glm.nb(riots ~ rgdpna_pc + pop +
               polity2, data=conflictData)
clarke_test(glm1b, glm2b)

## Ordered Logit: polr
library(MASS)
ol1 <- polr(as.factor(Amnesty) ~ log(rgdpna_pc) +
                  log(pop*1000) + polity2,
                data=conflictData)
ol2 <- polr(as.factor(Amnesty) ~ scale(rgdpna_pc) +
            scale(pop) + polity2,
            data=conflictData)
clarke_test(ol1, ol2)

## Ordered Logit: clm
library(ordinal)
ol1a <- clm(as.factor(Amnesty) ~ log(rgdpna_pc) +
              log(pop*1000) + polity2,
            data=conflictData)
ol2a <- clm(as.factor(Amnesty) ~ scale(rgdpna_pc) +
            scale(pop) + polity2,
            data=conflictData)
clarke_test(ol1a, ol2a)

## Multinomial Logit: multinom

library(nnet)
ml1 <- multinom(as.factor(Amnesty) ~ log(rgdpna_pc) +
              log(pop*1000) + polity2,
            data=conflictData)
ml2 <- multinom(as.factor(Amnesty) ~ scale(rgdpna_pc) +
              scale(pop) + polity2,
            data=conflictData)
clarke_test(ml1, ml2)


## Multinomial Logit: multinom

Conflict Data

Description

A country-year dataset containing information on conflict and other country attributes. These data come from multiple sources and are simply for the purposes of demonstrating how the functions in the package work. The data contain the following variables

Usage

conflictData

Format

A data frame with 4381 rows and 9 variables

GWNo

Gleditsch and Ward country number

Year

year

StateName

Country name

conflict_binary

Binary indicator of conflict

polity2

Polity IV indicator of regime type

Amnesty

Amnesty International Political Terror Scale Rating

riots

Number of riots in each country-yar

pop

Population in country (in thousands)

rgdpna_pc

PWT measure of GDP/capita

Calculate individual log-likelihood values

Description

Calculate individual log-likelihood values

Usage

indivLogLiks(model)

## S3 method for class 'glm'
indivLogLiks(model)

## S3 method for class 'lm'
indivLogLiks(model)

## S3 method for class 'orlm'
indivLogLiks(model)

## S3 method for class 'polr'
indivLogLiks(model)

## S3 method for class 'clm'
indivLogLiks(model)

## S3 method for class 'multinom'
indivLogLiks(model)

## S3 method for class 'negbin'
indivLogLiks(model)

Arguments

Details

The indivLogLiks function calls the appropriate method for calculating individual log likelihood values for the model. The function currently supports binomial, poisson and negative binomial GLMs, ordinal models estimated with either polr from the MASS package or clm from the ordinal package and multinomial models estimated with either multinom from the nnet package. Users can also write new methods for both indivLogLiks and nparams that would get called by the generic function.

For the purposes of the clarke_test function, the indivLogLiks functions are not intended to be called directly by the user.

Value

A vector of individual log-likelihood values for the model.

Find number of parameters in model

Description

Finds the number of parameters that were estimated in each model.

Usage

nparams(model)

## S3 method for class 'glm'
nparams(model)

## S3 method for class 'lm'
nparams(model)

## S3 method for class 'orlm'
nparams(model)

## S3 method for class 'polr'
nparams(model)

## S3 method for class 'clm'
nparams(model)

## S3 method for class 'multinom'
nparams(model)

## S3 method for class 'negbin'
nparams(model)

Arguments

Details

The function funds the number of parameters generally by counting the number of estimated parameters in the model's output.

For the purposes of the clarke_test function, the nparams functions are not intended to be called directly by the user.

Value

A scalar giving the number of parameters estimated in the model.

Print non-nested test results

Description

Prints results of the clarke_test function.

Usage

## S3 method for class 'nonnest.test'
print(x, digits = x$digits, ...)

Arguments

Value

Printed output that summarises the results of the clarke_test function.