R/isop.test.R

Defines functions isop.test

Documented in isop.test

## File Name: isop.test.R
## File Version: 0.20

############################################################
# testing the ISOP model
isop.test <- function( data, jackunits=20, weights=rep(1,nrow(data)) )
{
    dat <- as.matrix( data )
    dat.resp <- 1 - is.na( dat )
    dat[ is.na(dat) ] <- 0
    weights <- rep( 1, nrow(dat) )
    N <- nrow(dat)
    if ( length(jackunits==1)){
        jackunits <- ( 1:N ) %% jackunits
    }
    JJ <- max( jackunits )+1

    # apply Rcpp routine
    res <- isop_tests_cpp( dat=dat, dat.resp=dat.resp, weights=weights,
                jackunits=jackunits, JJ=JJ )

    #**********************
    # arrange output
    I <- ncol(dat)
    # Jackknife estimates
    itemstat <- as.data.frame( matrix( NA, nrow=I+1, ncol=3 ) )
    colnames(itemstat) <- c("parm", "est", "se" )
    W1test <- res$W1test
    itemstat[1,1] <- "test"
    itemstat[1,2] <- W1test[1]
    itemstat[1,3] <- sqrt( sum( ( W1test[-1] - mean( W1test[-1] ) )^2 ) * (JJ-1) / JJ )
    W1i <- res$W1i
    itemstat[-1,1] <- colnames(dat)
    itemstat[-1,2] <- W1i[,1]
    itemstat[-1,3] <- sqrt( rowSums( ( W1i[,-1] - rowMeans( W1i[,-1] ) )^2 ) * (JJ-1) / JJ )
    # include N's and p values
    itemstat <- data.frame("parm"=itemstat$parm, "N"=c(N, colSums( dat.resp * weights ) ),
            "M"=c(NA, colSums( dat.resp * weights * dat ) / colSums( dat.resp * weights ) ),
            itemstat[,-1] )
    itemstat$t <- itemstat$est / itemstat$se
    rownames(itemstat) <- NULL
    res <- list("itemstat"=itemstat, "Es"=res$Esi[,1], "Ed"=res$Edi[,1], JJ=JJ)
    class(res) <- "isop.test"
    return(res)
}
alexanderrobitzsch/sirt documentation built on March 18, 2024, 1:29 p.m.