R/helpers-bayesian-regression.R

In rdddr: Companion Datasets and Functions for Blair, Coppock, and Humphreys (2022) "Research Design: Declare, Diagnose, Redesign"

#' Tidy results from a stanreg regresion and exponentiate the estimated coefficient
#'
#' Note no standard errors or other summary statistics are provided
#'
#' See https://book.declaredesign.org/choosing-an-answer-strategy.html#bayesian-formalizations
#'
#' @param x A stanreg fit from stan_glm
#' @param conf.int Logical indicating whether or not to include a confidence interval in the tidied output. Defaults to FALSE.
#' @param conf.level The confidence level to use for the confidence interval if conf.int = TRUE. Must be strictly greater than 0 and less than 1. Defaults to 0.95, which corresponds to a 95 percent confidence interval.
#' @param exponentiate Logical indicating whether or not to exponentiate the the coefficient estimates. Defaults to FALSE. Note that standard errors are not included when \code{exponentiate = TRUE}.
#' @param ... Other arguments to broom.mixed::tidy
#'
#' @return data.frame of results
#'
#' @export
tidy_stan <- function(x, conf.int = FALSE, conf.level = 0.95, exponentiate = FALSE, ...) {
  if(!requireNamespace("broom.mixed")){
    message("The tidy_exponentiate function requires the 'broom.mixed' package.")
    return(invisible())
  }

  ret <- broom.mixed::tidy(x, conf.int = conf.int, conf.level = conf.level, ...)

  if (exponentiate) {
    ret$estimate <- exp(ret$estimate)
    if("conf.low" %in% colnames(ret)){
      ret$conf.low <- exp(ret$conf.low)
      ret$conf.high <- exp(ret$conf.high)
    }
    ret$std.error <- NULL
  }

  ret
}