R/intGen.R

In beckerbenj/eatAnalysis: Convenience function for various analysis types

####
#############################################################################
#' Generate interaction terms for \code{lm} formulas
#'
#' Specify two-way or three-way interactions from a set of independent variables.
#' The function might be useful, if the user has a regression model with more than
#' two predictors and want to specify all main effects plus all two-way interaction
#' effects but no higher-order interaction terms.
#'
#'@param vars Character vector of variables
#'@param upto The maximum way of interactions which should be created.
#'
#'@return A character string (snippet) useful for formula generation.
#'
#'@author Sebastian Weirich
#'
#'@examples
#' data(mtcars)
#' # relevant variables
#' vars <- c("cyl", "hp", "wt")
#' # cyl*hp*wt would specify two-way as well as three-way interactions
#' # generate formula snippet for two-way interactionn
#' frml <- intGen(vars = vars, upto = 2)
#' frml1<- as.formula(paste0("mpg ~ ", paste(vars, collapse=" + "), " + ",
#'         paste(frml, collapse=" + ")))
#' mod  <- lm(frml1, data = mtcars)
#'
#'@export
intGen <- function ( vars, upto = 3) {
  snip <- paste(lapply(2:upto, FUN = function ( v ) {
  spl <- data.frame ( combinat::combn(vars,v), stringsAsFactors = FALSE)
  sni <- paste(lapply(spl, FUN = function ( x ) {
    paste(x, collapse = " : ")}), collapse = " + ")
  return(sni)
  }), collapse = " + ")
return(snip)
}