CODEWITHSUNDEEP
rstatisticsformulaanova
Manuel Teodoro
Manuel Teodoro

Reputation: 151

Function to calculate ANOVA and writing basic stats in R

I am trying to write my first function in R, and I would appreciate some help.

As a researcher running experiments 2 or 3 times per year, I often have to compare the treatments using anova. I'm trying to automatize the Analysis of Variance to return a list with the results, a dataframe containing mean, n and sd, and a Tukey test to compare my treatments.

So far I have the following code:

# Function
Tt.aov <- function(vx, vtreat)
{
    VarAov <- aov(vx~vtreat)
    VarAnova <- anova(VarAov)
    p <- VarAnova$Pr[1]
    stats <- aggregate(vx, by=list (Treat=vtreat),
                       FUN=function(x) c(mean=mean(x), sd=sd(x), n=length(x)))
    #stats <- do.call(data.frame, stats)
    if (p<0.05) {Tuk <- TukeyHSD(VarAov)};
    aov.output <- list(anova=VarAnova,stats=stats,p=p)
    print(aov.output)
}

By now my function works very well for 1 treatment (vtreat). The question is, how can I keep my function open to add more treatments, let's say something like vx~vtraet1+vtreat2+vtreat3 or vx~vtreat1*vtreat2 while at the same time being able to use this parameters for the stats dataframe?

I could just do two or three functions that will be enough for my experiments, but I really want to understand and learn the process of using functions and formulas in R.

Upvotes: 1

Views: 170

Answers (1)

chamki
chamki

Reputation: 68

The problem isn't r's ability to take an undefined number of variables into a function (see here). It's the fact that aov is a wrapper for lm, which wants a formula. Fortunately, we can coerce a lot of things, like strings, into a formula. So, one way to get after what you want is to pass your variables as strings. The function will capture the first variable as vx and put the rest into a list represented by '...', which we will then have to break out into strings and put into the right format so we can coerce it into a formula.

data <- data.frame(dv=c(2,3,4,5),
                  iv1=c(10,11,13,16),
                  iv2=c(.3,.4,.2,.3))

Tt.aov <- function(vx, ...)
{
  trts <- list(...)
  temp = paste(vx, '~', trts[1])
  for (i in length(trts)-1) {
    temp <- append(temp, c('+', trts[i+1]))
  }
  form <- as.formula(paste(temp, collapse=''))
  VarAov <- aov(form)
  ...
}
Tt.aov('data$dv', 'data$iv1', 'data$iv2')

Upvotes: 1

Related Questions