R: population based on ratio or number
I am trying to combine some data that is both numeric and variable of factors. If the variable is numeric, I would like to have an average. If this is a factor, I would like to get the most frequent value. I wrote the following function, but I don't get the output I would like:
meanOrMostFreq <- function(x){
if(class(x) == 'factor'){
tbl <- as.data.frame(table(x))
tbl$Var1 <- as.character(tbl$Var1)
return(tbl[tbl$Freq == max(tbl$Freq),'Var1'][1])
}
if(class(x) == 'numeric'){
meanX <- mean(x, na.rm = TRUE)
return(meanX)
}
}
This is how I use it:
df1 <- iris[1:148,]
df1$letter1 <- as.factor(rep(letters[1:4], 37))
momf <- aggregate(.~ Species, df1, FUN = function(x) meanOrMostFreq(x))
And the results:
> momf
Species Sepal.Length Sepal.Width Petal.Length Petal.Width letter1
1 setosa 5.006000 3.428000 1.462000 0.246 2.46
2 versicolor 5.936000 2.770000 4.260000 1.326 2.54
3 virginica 6.610417 2.964583 5.564583 2.025 2.50
I am hoping to get the actual letter in the last column instead of a number. Any suggestions on what I am doing wrong?
Here's a way to use data.table
library(data.table)
setDT(df1)[ ,lapply(.SD, function(x) if(is.numeric(x)) mean(x, na.rm=TRUE) else
names(which.max(table(x)))) , by=Species]
# Species Sepal.Length Sepal.Width Petal.Length Petal.Width letter1
#1: setosa 5.006000 3.428000 1.462000 0.246 a
#2: versicolor 5.936000 2.770000 4.260000 1.326 c
#3: virginica 6.610417 2.964583 5.564583 2.025 a
Going through the interface of a formula to aggregate
, apparently, loses the metadata, which is its "factor"; this worked for me:
> meanOrMostFreq
function(x){
if(class(x) == 'factor'){
return( names(which.max(table(x))) )
}
if(class(x) == 'numeric'){
meanX <- mean(x, na.rm = TRUE)
return(meanX)
}
}
> aggregate(df1[-5], df1[5], meanOrMostFreq)
Species Sepal.Length Sepal.Width Petal.Length Petal.Width letter1
1 setosa 5.006000 3.428000 1.462000 0.246 a
2 versicolor 5.936000 2.770000 4.260000 1.326 c
3 virginica 6.610417 2.964583 5.564583 2.025 a
Since there are different behaviors for aggregate.formula
and aggregate.data.frame
, this seems like a mistake to me.
Alternative using package plyr
:
ddply(df1, .(Species), function(df) {
sapply(df, meanOrMostFreq)
})
[] 's