How to summarize over multiple columns programatically using ddply? - r

Question

Is it possible to specify which columns to aggregate over with ddply from the arguments of a function, without using eval + parse? Heres what I have so far:
x <- c(2,4,3,1,5,7)
y <- c(3,2,6,3,4,6)
group1 <- c("A","A","A","A","B","B")
group2 <- c("X","X","Y","Y","Z","X")
data <- data.frame(group1, group2, x, y)
Heres what I want the output to be:
aggFunction <- function(dataframe, toAverage, toGroup) {
out <- ddply(dataframe, toGroup, summarise,
x = mean(x),
y = mean(y))
return(out)
}
aggFunction(data, c("x", "y"), c("group1", "group2"))
# group1 group2 x y
# 1 A X 3 2.5
# 2 A Y 2 4.5
# 3 B X 7 6.0
# 4 B Z 5 4.0
Here's my solution using parse(eval):
aggFunction <- function(dataframe, toAverage, toGroup) {
toAverageArgs <- paste(toAverage, " = mean(", toAverage, ")", sep = "", collapse = ", ")
out <- eval(parse(text = paste("ddply(dataframe, toGroup, here(summarize),", toAverageArgs, ")")))
return(out)
}
This gives me the output that I want.
I'm wondering if there's a better way to do this. I'm aware of using do.call() and get(), but none of my attempts with these have worked.
Heres an attempt;
get(string) didn't work, but here(summarize) let me get the string value. Unfortunately, this means ddply treats them as strings:
aggFunction <- function(dataframe, toAverage, toGroup) {
string <- paste(toAverage, " = mean(", toAverage, ")", sep = "", collapse = ", ")
out <- ddply(dataframe, toGroup, here(summarise), string)
return(out)
}
aggFunction(data, c("x", "y"), c("group1", "group2"))
# group1 group2 ..2
# 1 A X x = mean(x), y = mean(y)
# 2 A Y x = mean(x), y = mean(y)
# 3 B X x = mean(x), y = mean(y)
# 4 B Z x = mean(x), y = mean(y)
Also tried do.call, but they were still treated as strings:
aggFunction <- function(dataframe, toAverage, toGroup) {
string <- paste(toAverage, " = mean(", toAverage, ")", sep = "", collapse = ", ")
print(string)
args <- list(dataframe, toGroup, here(summarise), string)
out <- do.call(ddply, args)
return(out)
}
aggFunction(data, c("x", "y"), c("group1", "group2"))
# group1 group2 "x = mean(x), y = mean(y)"
# 1 A X x = mean(x), y = mean(y)
# 2 A Y x = mean(x), y = mean(y)
# 3 B X x = mean(x), y = mean(y)
# 4 B Z x = mean(x), y = mean(y)
Lastly I tried hardcoding in mean(), but then I couldnt set the column name. If I use get(testVar) = mean(get(testVar)) I get unexpected =.
aggFunction <- function(dataframe, toAverage, toGroup) {
testVar <- "x"
out <- ddply(dataframe, toGroup, here(summarise),
get(testVar) = mean(get(testVar))
##
return(out)
}

Answer 1

Using aggregate in base R
aggFunction <- function(dataframe, toAverage, toGroup) {
aggregate(dataframe[, toAverage], dataframe[, toGroup], mean)
}
aggFunction(data, c("x", "y"), c("group1", "group2"))
group1 group2 x y
1 A X 3 2.5
2 B X 7 6.0
3 A Y 2 4.5
4 B Z 5 4.0

Answer 2

You can consider dplyr package - generally it's much faster than plyr and also has pretty syntax.
library(dplyr)
x <- c(2,4,3,1,5,7)
y <- c(3,2,6,3,4,6)
group1 <- c("A","A","A","A","B","B")
group2 <- c("X","X","Y","Y","Z","X")
aggFunction <- function(dataframe, toAverage, toGroup) {
dataframe %>%
group_by_(.dots = toGroup) %>%
summarise_(.dots = setNames(sprintf("mean(%s)", toAverage), toAverage))
}
data <- data.frame(group1, group2, x, y)
aggFunction(data, c("x", "y"), c("group1", "group2"))
It gives:
group1 group2 x y
1 A X 3 2.5
2 A Y 2 4.5
3 B X 7 6.0
4 B Z 5 4.0

Answer 3

This is a lot easier if you melt the data frame first, do the calculations when in long format, and then cast back.
library(reshape2)
library(plyr)
aggFunction <- function(d1, toAverage, toGroup) {
d2 <- melt(d1, id.vars=toGroup, measure.vars=toAverage)
d3 <- ddply(d2, ~group1 + group2 + variable, summarize, mean=mean(value))
dcast(d3, group1 + group2 ~ variable, value.var="mean")
}
aggFunction(data, c("x", "y"), c("group1", "group2"))
## group1 group2 x y
## 1 A X 3 2.5
## 2 A Y 2 4.5
## 3 B X 7 6.0
## 4 B Z 5 4.0