R/step_lm.R

In Blockhead-yj/report.iqz: Make Potential Predction Report of IQUIZOO

#' @title Select predictor games by step regression
#'
#' @description Usually users will have lots of iquizoo games test at once,
#' however it is bad idea to put all the games into a model to predict academic
#' score, which might cause the problem of overfitting or a decline of adjust r squared.
#' So we choose a subset of iquizoo games to be placed into the model by step regression.
#'
#' @param data_wider Data.frame, which each row contains all the observations of a single person
#' @param formula Formula, indicate which variables should be included in model
#' @param n_tasks, Numerical, the least number of tasks that a step method should be used
#' @param ... Other arguments used in stats::step function
#'
#' @return A lm object
#'
#' @export

step_lm <- function(data_wider, formula, n_tasks=5, trace = FALSE, ...){
  complete_data <- tidyr::drop_na(data_wider)
  tmp_model <- lm(formula, data = complete_data)
  if(length(tmp_model$terms |> attr("term.labels")) >= n_tasks)
    tmp_model <- step(tmp_model, trace = trace, ...)

  tmp_model
}