R/constraints.R

In REMLA: Robust Expectation-Maximization Estimation for Latent Variable Models

#' CFA Constraints Matrix
#' @description
#' Function to create constraints matrix based on user inputs for Confirmatory Factor Analysis
#' @param X numeric data frame containing all the variables needed in the factor analysis
#' @param order original order of variables in the data frame
#' @returns matrix of zeros and ones indicating the relationship that wants to be modeled
#' @author Bryan Ortiz-Torres (bortiztorres@wisc.edu); Kenneth Nieser (nieser@stanford.edu)
#' @references Nieser, K. J., & Cochran, A. L. (2023). Addressing heterogeneous populations in latent variable settings through robust estimation. Psychological methods, 28(1), 39.
#' @noRd


constraints <- function(X,order) {

  lines <- strsplit(X, "\n")[[1]]
  indicators <- list()
  latent_vars <- c()

  for (line in lines) {
    if (grepl("=~", line)) {
      parts <- strsplit(line, "=~")[[1]]
      latent_var <- trimws(parts[1])
      indicators <- c(indicators, strsplit(trimws(parts[2]), "\\+")[[1]]) #change this part to allow for not spaces between plus signs
      indicators <- trimws(indicators)
      latent_vars <- c(latent_vars, latent_var)
    }
  }
  unique_indicators <- unique(indicators)
  if(length(unique_indicators) != length(order)) stop(paste0("The number of variables in the dataset (", length(order), ") does not match the number of variables in the model (", length(unique_indicators), ")."))

  matrix_data <- matrix(0, nrow = length(unique_indicators), ncol = length(latent_vars))
  rownames(matrix_data) <- unique_indicators
  colnames(matrix_data) <- latent_vars

  for (line in lines) {
    if (grepl("=~", line)) {
      parts <- strsplit(line, "=~")[[1]]
      latent_var <- trimws(parts[1])
      indicators <- strsplit(trimws(parts[2]), "\\+")[[1]]
      indicators <- trimws(indicators)

      for (indicator in indicators) {
        matrix_data[indicator, latent_var] <- 1
      }
    }
  }

  # re-order so that variables are in the same order as in the dataset
  matrix_data2 = matrix_data[order,]

  return(matrix_data2)
}