R/anova_immer.R

In immer: Item Response Models for Multiple Ratings

## File Name: anova_immer.R
## File Version: 0.22

##############################################################
# anova method immer_cml
anova_immer <- function( object, ... )
{
    cl2 <- paste(match.call())[-1]
    if (length(list(object, ...)) !=2){
        stop("anova method can only be applied for comparison of two models.\n")
    }
    objects <- list(object, ...)
    model1 <- objects[[1]]
    model2 <- objects[[2]]

    # define some necessary parameters
    model1$loglike <- model1$loglike
    model1$Npars <- model1$npars
    model2$loglike <- model2$loglike
    model2$Npars <- model2$npars
    # test
    dfr1 <- data.frame( "Model"=cl2[1],
                "loglike"=model1$loglike,
                "Deviance"=-2*model1$loglike )
    dfr1$Npars <- sum(model1$Npars)
    dfr2 <- data.frame( "Model"=cl2[2],
                    "loglike"=model2$loglike, "Deviance"=-2*model2$loglike )
    dfr2$Npars <- sum(model2$Npars)
    dfr <- rbind( dfr1, dfr2 )
    dfr <- dfr[ order( dfr$Npars ), ]
    dfr$Chisq <- NA
    dfr$df <- NA
    dfr$p <- NA
    dfr[1,"Chisq"] <- dfr[1,"Deviance"] - dfr[2,"Deviance"]
    dfr[1,"df"] <- abs( dfr[1,"Npars"] - dfr[2,"Npars"] )
    dfr[ 1, "p" ] <- round( 1 - stats::pchisq( dfr[1,"Chisq"], df=dfr[1,"df"] ), 5 )
    for ( vv in 2:( ncol(dfr))){ dfr[,vv] <- round( dfr[,vv], 5 ) }
    print( dfr )
    invisible(dfr)
}
##############################################################

anova.immer_cml <- anova_immer
anova.immer_hrm <- anova_immer
anova.lc2_agreement <- anova_immer
anova.immer_latent_regression <- anova_immer