R/scale_parameters.R

In petedodd/MCIR: Monte Carlo Inference Routines: A Collection of Generic Algorithms for Bayesian Inference

scale_parameters <- function(prior, params){
    ## function scaled = scale_parameters(prior, params)
    ## % scaled = scale_parameters(prior, params)
    ## %
    ## % This function will scale parameters based on their priors. If a prior is
    ## % uniform over a range then the parameter will be scaled, such that the
    ## % range covers 0->1. If a prior is Gaussian then the parameter will be
    ## % scaled such that it will be a Gaussian with zero mean and unit variance.

    lp <- ncol(params)                  #pjd: use df for params
    scaled <- matrix(NA,nrow=nrow(params),ncol=ncol(params)) #pjd - vectorising

    for( i in 1:lp ) {
        priortype <- prior[i,2]
        p3 <- prior[i,3]
        p4 <- prior[i,4]

        ## % currently only handles uniform or Gaussian priors
        if( priortype == 'uniform' )
            scaled[,i] <- (params[,i] - p3)/(p4 - p3)
        if( priortype == 'gaussian' )
            scaled[,i] <- (params[,i] - p3)/p4            
        if( priortype == 'jeffreys' )
            scaled[,i] <- (log10(params[,i]) - log10(p3))/(log10(p4) - log10(p3)) 
    }

    return(scaled) 
}