R/param.logit.blme.R

In bensoltoff/ZeligBLME: A Zelig Model

#' Extract Samples from a Distribution in Order to Pass Them to the \code{qi} Function
#' (this is primarily a helper function for the logit.blme model)
#' @param obj a zelig object
#' @param num an integer specifying the number of simulations to compute
#' @param ... additional parameters
#' @return a list specifying link, link-inverse, random samples, and ancillary parameters
#' @export
param.logit.blme <- function(obj, num=1000, ...) {
  zelig <- obj
  # 
  obj <- GetObject(obj)
  
  # this is kludge, while functions are not being forwarded correctly
  # sigma <- getMethod('sigma', 'mer') Approach stopped working
  
  fixed <- fixef(obj)
  vars <- ranef(obj, postVars=TRUE)
  corr <- VarCorr(obj)
  gammas <- NULL
  
  for (key in names(vars)) {
    # Sigma makes sense, but why is mu the zero vector?
    Sigma <- corr[[key]]
    mu <- rep(0, ncol(Sigma))
    
    #
    gammas[[key]] <- mvrnorm(num, mu, Sigma)
  }
  
  names(gammas) <- names(vars)
  betas <- mvrnorm(num, fixef(.fitted), vcov(.fitted))
  scale <- lme4::sigma(.fitted)
  
  list(
    coef = betas,
    alpha = list(scale = scale, gammas = gammas),
    fam = eval(zelig$call$family, zelig$env)
  )
}