R/chisq.R

In oottest: Out-of-Treatment Testing

#' Chi-Square test for a single theory
#' @param data matrix rows: choices, columns: treatments
#' @param prediction list of matrices each in the same format as above
#'
#' @return Chi-square test for a single theory
#'
#' @noRd
get_chi_sq <- function(data, prediction) {
  num_treatments <- ncol(data)
  chi_sq <- 0
  for (i in 1:num_treatments) {
    chi_sq <- chi_sq + stats::chisq.test(data[, i], p = prediction[, i])$statistic
  }
  p_value <- stats::pchisq(chi_sq, df = num_treatments, lower.tail = FALSE)
  output <- c(chi_sq, p_value)
  names(output) <- c("chi-sq", "p-value")
  return(output)
}


#' Chi-Square Tests for multiple theories
#'
#' @param data Data of the experiment (matrix of ints/data.frame): columns are observations of action 1,2,3..., rows are treatments.
#' @param predictions Model prediction for one theory: rows/cols as data
#'
#' @return Output of all chi square tests
#'
#' @noRd
get_all_chi_sq <- function(data, predictions) {
  output <- c()
  num_theories <- dim(predictions)[3]
  for (i in 1:num_theories) {
    output <- rbind(output, get_chi_sq(data, predictions[, , i]))
  }
  rownames(output) <- colnames(predictions[1, , ])
  output[, "chi-sq"] <- round(output[, "chi-sq"], 2)
  output[, "p-value"] <- round(output[, "p-value"], 12)
  return(output)
}