CODEWITHSUNDEEP
rfor-loopanova
Giffredo
Giffredo

Reputation: 79

ANOVA repeated measure on multiple data frames r

I have hundreds of data frames. I need to perform ANOVA RM tests on each of these data frames. The output should be one single data frame with the mean of each p-value.

I tried:

#crate dataframes
df1 <- data.frame(replicate(16,sample(-10:10,10,rep=TRUE)))
df2 <- data.frame(replicate(16,sample(-10:10,10,rep=TRUE)))
df3 <- data.frame(replicate(16,sample(-10:10,10,rep=TRUE)))
Group <- c(rep("A",8),rep("B",8))
Time <- c(rep("before",4),rep("after",4),rep("before",4),rep("after",4))
Name <- rep(rep(1:4, 4))
conds <- data.frame(Name,Time,Group)
#create list
list <- list(df1,df2,df3)

#for loop ANOVA repeated measures
for ( i in list){
  data <- cbind(conds,i)
  t=NULL
  name <- colnames(data)[4:ncol(data)]
  for(i in 4:ncol(data)) { z <- aov(data[,i] ~ Group*Time+Error(Name/(Group*Time)), data=data) 
  sz <- as.list(summary(z))
  t <- as.data.frame(c(t,sz[4]$`Error: Name:Group:Time`[[1]]$`Pr(>F)`[1]))
  t
  }
}
mean(t)

Upvotes: 1

Views: 504

Answers (1)

jay.sf
jay.sf

Reputation: 72583

R as a vectorized language is designed to avoid for loops where possible. You could do an sapply approach.

When you list your data frames use names like df1=, which later helps in the result on which of them were done calculations.

(And don't use list as object name since you'll get confused because there is also a list function. Also data, df and friends are "bad" names, you may always check, using e.g. ?list if the name is already occupied.)

list1 <- list(df1=df1, df2=df2, df3=df3)

res <- sapply(list1, function(x) {
  dat <- cbind(conds, x)
  sapply(dat[-(1:3)], function(y) {
    z <- aov(y ~ Group*Time + Error(Name/(Group*Time)), data=dat)
    sz <- summary(z)
    p <- sz$`Error: Name:Group:Time`[[1]][1, 5]
    p
  })
})

From the resulting matrix we take the column means.

colMeans(res)
#       df1       df2       df3 
# 0.4487419 0.4806528 0.4847789

Data:

set.seed(42)
df1 <- data.frame(replicate(16,sample(-10:10,16,rep=TRUE)))
df2 <- data.frame(replicate(16,sample(-10:10,16,rep=TRUE)))
df3 <- data.frame(replicate(16,sample(-10:10,16,rep=TRUE)))
conds <- data.frame(Name=c(rep("A",8),rep("B",8)), 
                    Time=c(rep("before",4),rep("after",4),
                           rep("before",4),rep("after",4)),
                    Group=rep(1:4, 4))

Upvotes: 1

Related Questions