R/frm_prepare_data_include_latent_data.R

In mdmb: Model Based Treatment of Missing Data

## File Name: frm_prepare_data_include_latent_data.R
## File Version: 0.224


##################################################################
# include latent data according nodes for
# calculation in EM algorithm
frm_prepare_data_include_latent_data <- function( dat, var_mm,
    nodes_mm, ind_mm )
{
    resp_varname <- paste0("resp_", var_mm)
    dat[, resp_varname ] <- 1 - is.na(dat[,var_mm] )
    dat$resp_all <- dat$resp_all * dat[, resp_varname ]
    dat1 <- dat[ dat[, resp_varname ]==1, ]
    dat0 <- dat[ dat[, resp_varname ]==0, ]
    N0 <- nrow(dat0)
    NX <- length(nodes_mm)
    dat12 <- dat1
    # if (N0>0){
    #    for (xx in 1:NX){
    #        # xx <- 1
    #        dat11 <- dat0
    #        dat11[,var_mm] <- nodes_mm[xx]
    #        dat11$weights <- dat11$weights * ind_mm$nodes_weights[xx]
    #        dat12 <- rbind( dat12, dat11 )
    #    }
    # }
    if (N0>0){
        rep_ind <- rep( 1:N0, NX )
        dat11 <- dat0[ rep_ind,  ]
        dat11[,var_mm] <- rep( nodes_mm, each=N0 )
        dat11$weights <- dat11$weights * rep( ind_mm$nodes_weights, each=N0 )
        #*** include delta_nodes for some models
        model_mm <- ind_mm$model
        models0 <- c("linreg", "yjtreg", "bctreg")
        if (sum( model_mm %in% models0 )> 0 ){
            d1 <- cbind( c( diff(nodes_mm), NA), c(NA, diff(nodes_mm) ) )
            diff_nodes <- rowMeans(d1, na.rm=TRUE)
            dat11$delta_nodes <- dat11$delta_nodes * rep( diff_nodes, each=N0 )
        }
        dat12 <- rbind( dat12, dat11 )
    }
    dat12 <- dat12[ order(dat12$case), ]
    return(dat12)
}