R/anova_effect.R

In McCartneyAC/mccrr: Personal R Package

#' Calculate Omega Squared for a Univariate ANOVA
#'
#' Credit to J. Patrick Meyer
#'
#' @param aovm An ANOVA model
#'
#'
#' @export
omega_sq <- function(aovm){
  sum_stats <- summary(aovm)[[1]]
  SSm <- sum_stats[["Sum Sq"]][1]
  SSr <- sum_stats[["Sum Sq"]][2]
  DFm <- sum_stats[["Df"]][1]
  MSr <- sum_stats[["Mean Sq"]][2]
  W2 <- (SSm-DFm*MSr)/(SSm+SSr+MSr)
  return(W2)
}


#' Calculate Eta Squared for a Univariate ANOVA
#'
#' Adapted from J. Patrick Meyer
#'
#' @param aovm An ANOVA model
#'
#' @export
eta_sq<-function(aovm){
  sum_stats <- summary(aovm)[[1]]
  SSm <- sum_stats[["Sum Sq"]][1]
  SSr <- sum_stats[["Sum Sq"]][2]
  e2 <- (SSm)/(SSm+SSr)
  return(e2)
}


#' Calculate D and F family Effect Sizes for a Univariate ANOVA
#'
#' Adapted from J. Patrick Meyer
#'
#' @param aovm An ANOVA model
#' @param continuous That model's continuous outcome
#' @param group That model's grouping predictor variable
#'
anova_effect<-function(aovm, continuous, group){
  sum_stats <- summary(aovm)[[1]]
  SSm <- sum_stats[["Sum Sq"]][1]
  SSr <- sum_stats[["Sum Sq"]][2]
  DFm <- sum_stats[["Df"]][1]
  MSr <- sum_stats[["Mean Sq"]][2]
  w2 <- (SSm-DFm*MSr)/(SSm+SSr+MSr)
  e2 <- (SSm)/(SSm+SSr)
  grandmean<-mean(continuous)
  groupmeans<-aggregate(continuous, list(group), mean)[,2]
  difmeans<-sum((groupmeans-grandmean)^2)
  d<-sqrt(1/(DFm)*(difmeans/MSr))
  f<-sqrt(1/(DFm+1)*(difmeans/MSr))
  mat<-cbind(d,f,e2,w2)
  colnames(mat) <- c("d effect size","f effect size","eta squared","omega squared")
  return(mat)
}