Question

I am wondering what would be the best way to add, for example, columns of quantiles to the dataset. I was thinking to use ave() function for that, something like ave(iris$Sepal.Length, iris$Species, FUN=quantile) – but in this case ave() merges values returned by quantile() (which in this case returns 5 values per subset) and cut them for the length of iris…

Thanks in advance for suggestions!

Answer 1

There are a lot of SO questions on this general topic, recommending various uses of ave(), aggregate(), plyr(), reshape2::cast, or data.table depending on personal preference, readability, compactness, flexibility, speed … Here’s a simple solution with aggregate() that seems to do what you want:

(aa <- aggregate(Sepal.Length~Species,data=iris,quantile))

##      Species Sepal.Length.0% Sepal.Length.25% Sepal.Length.50% Sepal.Length.75%
## 1     setosa           4.300            4.800            5.000            5.200
## 2 versicolor           4.900            5.600            5.900            6.300
## 3  virginica           4.900            6.225            6.500            6.900
##   Sepal.Length.100%
## 1             5.800
## 2             7.000
## 3             7.900

edit: re-reading/based on comment, this is not what you want: you need the summarized values replicated for each row, not just once per group.

Perhaps

merge(iris,aa,by="Species")

although that gives a slightly weird data frame (the last “column” is actually a matrix).

It’s a little bit magical, but

merge(iris,with(aa,data.frame(Species,Sepal.Length)))

is better — it unpacks the weird data frame returned by aggregate() a bit more (the names are still a bit wonky).