R/MHprop_hrm.R

In immer: Item Response Models for Multiple Ratings

## File Name: MHprop_hrm.R
## File Version: 0.142


######################################################################
MHprop_hrm <- function( MHprop, b, a, phi, theta, iter, burnin )
{

    psi <- phi
    MHprop <- list()

    # logical indicating whether standard deviation of the
    # proposal distribution is calculated according to
    # Browne and Draper (2006)
    MHprop$refresh_formula <- TRUE

    MHprop$SD <- list( 'b'=.4+0*b, 'a'=.4 +0*a, 'phi'=.2 + 0*phi, 'psi'=.4 + 0*psi,
            'theta'=.5 + 0*theta )
    MHprop$accept <- list('b'=0 + 0*b, 'a'=0 + 0*a, 'phi'=0+0*phi,
                        'psi'=0 + 0*psi, 'theta'=0 + 0*theta  )

    MHprop$refresh_count$b <- 0
    MHprop$refresh_count$a <- 0
    MHprop$refresh_count$phi <- 0
    MHprop$refresh_count$psi <- 0
    MHprop$refresh_count$theta <- 0

    RI <- 25
    MHprop$refresh_iter$b <- RI
    MHprop$refresh_iter$a <- RI
    MHprop$refresh_iter$phi <- RI
    MHprop$refresh_iter$psi <- RI
    MHprop$refresh_iter$theta <- RI

    MHprop$refresh_SDchange$b <- .02
    MHprop$refresh_SDchange$a <- .02
    MHprop$refresh_SDchange$phi <- .02
    MHprop$refresh_SDchange$psi <- .02
    MHprop$refresh_SDchange$theta <- .05


    vars <- c('b','a', 'phi', 'psi', 'theta')

    # compute iterations for which MH updatings must be computed
    refresh_iters <- sort( unique( unlist( MHprop$refresh_iter ) ) )
    RI <- length(refresh_iters)
    v1 <- NULL
    for (rr in 1L:RI){
        l1 <- (1L:iter %% refresh_iters[rr] )==0
        l1 <- (1L:iter)[l1]
        v1 <- c( v1, l1 )
    }
    v2 <- sort( unique(v1) )
    v2 <- v2[ v2 <=burnin ]
    MHprop$ITER_refreshing <- v2

    # refreshing variables
    MHprop$VARS_refreshing <- c('b', 'phi', 'psi', 'theta')

    # boundaries for acceptance rates
    MHprop$accept_bounds <- c( .4, .6 )

    return(MHprop)
}
######################################################################