Overall ANOVA Table in R

[Krishnaa]

Looks like this problem was never solved.

ANOVA Table for Full Linear Model

I'm teaching Multiple Regression again this semester, so I'm once again reminded that the ANOVA output from R is not the same as from other statistical software. A quick google turned up this post from r-help which has essentially my exact question, so I'll quote from it here (data changed for better reproducibility): I am trying to figure out how to get an ANOVA table that shows the sum of squares. degrees of freedom, etc, for the full model versus the error (aka residuals). Here is an example of the kind of table I'd like to get: Analysis of Variance Source DF SS MS F P Regression 2 0.9715 0.485728 52.13 0.000 Error…

This solution might help?

anova_table <- function(regression)
{
  
  regress_1 <- anova(regression)
  #regress_1
  
  anovatablereg <- regress_1[-nrow(regress_1),]
  #anovatablereg
  
  ANOVA_table_Total = data.frame(DF=sum(anovatablereg$Df), SS=sum(anovatablereg$`Sum Sq`),MS=sum(anovatablereg$`Sum Sq`)/sum(anovatablereg$Df) , F_Value=NA, P_Value=NA)
  ANOVA_table_Total[2,] = regress_1[nrow(regress_1),]
  ANOVA_table_Total[3,] = ANOVA_table_Total[1,] + ANOVA_table_Total[2,]
  ANOVA_table_Total[1,4] = ANOVA_table_Total[1,3]/ANOVA_table_Total[2,3]
  ANOVA_table_Total[1,5] = 1- pf(ANOVA_table_Total[1,4],ANOVA_table_Total[1,1],ANOVA_table_Total[2,1])
  ANOVA_table_Total[3,3] = NA
  
  rownames(ANOVA_table_Total) = c("REGRESSION", "ERROR","TOTAL")
  
  return(ANOVA_table_Total)
  
}

#regression is an lm_object
#i.e when you assign x <- lm(), X is the regression. 
#to use the function do this
#define your lm model
#then write this code
#variable_name <- anova_table(x)
#now variable_name is a data_frame containing the anova table for the regression 
#print() or View() to see the table.

I am a novice R user. So be easy on the criticism. I took advantage of the lm() and anova() functions to get the overall ANOVA table. Hope this helps!

Thanks.

Krishnaa.

[technocrat]

Here's the reprex from this post

anova_alt = function (object, reg_collapse=TRUE,...) 
{
  if (length(list(object, ...)) > 1L) 
    return(anova.lmlist(object, ...))
  if (!inherits(object, "lm")) 
    warning("calling anova.lm(<fake-lm-object>) ...")
  w <- object$weights
  ssr <- sum(if (is.null(w)) object$residuals^2 else w * object$residuals^2)
  mss <- sum(if (is.null(w)) object$fitted.values^2 else w * 
               object$fitted.values^2)
  if (ssr < 1e-10 * mss) 
    warning("ANOVA F-tests on an essentially perfect fit are unreliable")
  dfr <- df.residual(object)
  p <- object$rank
  if (p > 0L) {
    p1 <- 1L:p
    comp <- object$effects[p1]
    asgn <- object$assign[stats:::qr.lm(object)$pivot][p1]
    nmeffects <- c("(Intercept)", attr(object$terms, "term.labels"))
    tlabels <- nmeffects[1 + unique(asgn)]
    ss <- c(vapply(split(comp^2, asgn), sum, 1), ssr)
    df <- c(lengths(split(asgn, asgn)), dfr)
    if(reg_collapse){
      if(attr(object$terms, "intercept")){
        collapse_p<-2:(length(ss)-1)
        ss<-c(ss[1],sum(ss[collapse_p]),ss[length(ss)])
        df<-c(df[1],sum(df[collapse_p]),df[length(df)])
        tlabels<-c(tlabels[1],"Source")
      } else{
        collapse_p<-1:(length(ss)-1)
        ss<-c(sum(ss[collapse_p]),ss[length(ss)])
        df<-c(df[1],sum(df[collapse_p]),df[length(df)])
        tlabels<-c("Regression")
      }
    }
  }else {
    ss <- ssr
    df <- dfr
    tlabels <- character()
    if(reg_collapse){
      collapse_p<-1:(length(ss)-1)
      ss<-c(sum(ss[collapse_p]),ss[length(ss)])
      df<-c(df[1],sum(df[collapse_p]),df[length(df)])
    }
  }
  
  ms <- ss/df
  f <- ms/(ssr/dfr)
  P <- pf(f, df, dfr, lower.tail = FALSE)
  table <- data.frame(df, ss, ms, f, P)
  table <- rbind(table, 
                 colSums(table))
  table$ms[nrow(table)]<-table$ss[nrow(table)]/table$df[nrow(table)]
  table[length(P):(length(P)+1), 4:5] <- NA
  dimnames(table) <- list(c(tlabels, "Error","Total"), 
                          c("Df","SS", "MS", "F", 
                            "P"))
  if (attr(object$terms, "intercept")){
    table <- table[-1, ]
    table$MS[nrow(table)]<-table$MS[nrow(table)]*(table$Df[nrow(table)])/(table$Df[nrow(table)]-1)
    table$Df[nrow(table)]<-table$Df[nrow(table)]-1
  }
  structure(table, heading = c("Analysis of Variance Table\n"), 
            class = c("anova", "data.frame"))
}

## Warpbreaks example
fm1 <- lm(breaks ~ wool*tension, data = warpbreaks)
anova_alt(fm1)
#> Analysis of Variance Table
#> 
#>        Df    SS     MS      F         P
#> Source  5  3488 697.54 5.8279 0.0002772
#> Error  48  5745 119.69                 
#> Total  53 52018 981.47