R/mdmb_regression_loglike_case.R

In alexanderrobitzsch/mdmb: Model Based Treatment of Missing Data

## File Name: mdmb_regression_loglike_case.R
## File Version: 0.201

#**** evaluate individual likelihood
mdmb_regression_loglike_case <- function(y, linear.predictor,
    fitted.values, type, beta, df, index_beta=NULL, index_thresh=NULL )
{
    np <- length(beta)
    #**********************
    # logistic regression
    if (type=='logistic'){
        loglike_case <- ifelse( y==1, fitted.values, 1 - fitted.values )
    }
    #**********************
    # ordinal probit model
    if (type=='oprobit'){
        thresh <- logthresh_2_thresh(x=beta[ index_thresh ])
        loglike_case <- mdmb_regression_oprobit_density( y=y, ypred=linear.predictor,
                            thresh=thresh, log=FALSE )
    }

    #**********************
    # yjt regression
    if (type=='yjt'){
        sigma <- beta[ np-1 ]
        lambda <- beta[ np ]
        loglike_case <- dyjt_scaled( y, location=linear.predictor, shape=sigma,
                            lambda=lambda, df=df )
    }
    #**********************
    # bct regression
    if (type=='bct'){
        sigma <- beta[ np-1 ]
        lambda <- beta[ np ]
        loglike_case <- dbct_scaled( y, location=linear.predictor, shape=sigma,
                            lambda=lambda, df=df )
    }
    #------------------------------------
    #--- log-likelihood
    eps <- 1E-50
    loglike_case <- log( loglike_case + eps )
    #------------------------------------
    return(loglike_case)
}