Calculate Bivariate Residuals (BVRs) for latent class models in R (poLCA)

bvr_polca.R

# Author: Daniel Oberski
# Date: 2017-08-01

# Bivariate residual statistic for latent class analysis
# Calculate the BVR for poLCA objects

# Argument: a poLCA object
# Value: a dist object with BVRs
# Example: bvr(fit)

bvr <- function(fit) {
  stopifnot(class(fit) == "poLCA")
  
  ov_names <- names(fit$predcell)[1:(ncol(fit$predcell) - 2)]
  ov_combn <- combn(ov_names, 2)
  
  get_bvr <- function(ov_pair) {
    form_obs <- as.formula(paste0("observed ~ ", ov_pair[1], " + ", ov_pair[2]))
    form_exp <- as.formula(paste0("expected ~ ", ov_pair[1], " + ", ov_pair[2]))
    
    counts_obs <- xtabs(form_obs, data = fit$predcell)
    counts_exp <- xtabs(form_exp, data = fit$predcell)
    
    bvr <- sum((counts_obs - counts_exp)^2 / counts_exp)
    
    bvr
  }
  
  bvr_pairs <- apply(ov_combn, 2, get_bvr)
  # names(bvr_pairs) <- apply(ov_combn, 2, paste, collapse = "<->")
  attr(bvr_pairs, "class") <- "dist"
  attr(bvr_pairs, "Size") <- length(ov_names)
  attr(bvr_pairs, "Labels") <- ov_names
  attr(bvr_pairs, "Diag") <- FALSE
  attr(bvr_pairs, "Upper") <- FALSE
  
  bvr_pairs
}

When I use it I get a matrix of NaNs. I'm investigating this problem, and is mainly caused in the line 24. I get many Infinite values. I don't know if it is useful to introduce: bvr[is.infinite(bvr)]<-NA; bvr <- sum(bvr,na.rm=T)