Descriptive Statistics by Group for multiple variables

Question

Given the data frame (df)

Hup Hop testA   testB
Y   Hi  1   1
N   Lo  2   2
Y   Mi  3   3
N   No  4   4
Y   Hi  5   5
N   Lo  6   6
Y   Mi  7   7
N   No  8   8
Y   Hi  9   9
N   Lo  10  10
Y   Mi  11  11
N   No  12  12

I want the descriptive statistics (mean and sd) of testA and testB for the grouping variables Hup and Hop. I want to something like this.

hup testA.mean  testA.sd    testB.mean  testB.sd
y   7            3.742            7      3.742
n   6            3.742            6      3.742
hop testA.mean  testA.sd    testB.mean  testB.sd
hi  etc           Etc            etc       Etc
lo  etc           Etc            etc       Etc
mi  etc           Etc            etc       Etc

Using e.g., ddply(df,~hup,summarise,mean=round(mean(testA),3),sd=round(sd(testA),3)) would solve a part of the problem. But I want to speed up the process: learn how to use R. So, I thought:

lapply(df[ , c("testA", "testB")], function(x){ ddply(df, ~df[ , c("hup")], function(x) {mean(x)} )})

which is not working, it returns NA’s, misses the SD and reports only results for hup.

Q: How to produce descriptive statistics for several groups with multiple variables?

Check out `aggregate(testA~Hup+Hop, df, mean)` – Julián Urbano Feb 27 '14 at 11:03 — Julián Urbano, Feb 27 '14 at 11:03

adibender · Accepted Answer · 2014-02-27T11:50:19.530

3

For display I think tabular function from tables package is easiest:

library(tables)
tabular(Hup + Hop ~ (testA + testB)*((n = 1) + mean + sd), data = df)
##       testA         testB        
##       mean  sd    n mean  sd    n
##Hup N  7     3.742 6 7     3.742 6
##    Y  6     3.742 6 6     3.742 6
##Hop Hi 5     4.000 3 5     4.000 3
##    Lo 6     4.000 3 6     4.000 3
##    Mi 7     4.000 3 7     4.000 3
##    No 8     4.000 3 8     4.000 3

U can also wrap the tabular() object in latex() to output the table in LaTeX syntax.

edited Feb 27 '14 at 11:50

answered Feb 27 '14 at 11:26

adibender

7,288
3
37
41

Thank you @adibender, it seems that `tables` is the package I have looked for. Finding the right (best) package to solve a problem is hard in the overwhelming offer of packages. Thanks! – Adam Feb 27 '14 at 11:51

Jake Burkhead · Answer 2 · 2014-02-27T11:22:11.050

library(reshape2)
library(plyr)

dfm <- melt(df, id.vars = c("Hup", "Hop"))

splits <- list(.(Hup, variable), .(Hop, variable))
## or something like
## splits <- data.frame(rbind(head(names(dfm), -2), "variable")) 
lapply(splits, ddply, .data = dfm, .fun = summarize, mean = mean(value), sd = sd(value))

## [[1]]
##   Hup variable mean       sd
## 1   N    testA    7 3.741657
## 2   N    testB    7 3.741657
## 3   Y    testA    6 3.741657
## 4   Y    testB    6 3.741657

## [[2]]
##   Hop variable mean sd
## 1  Hi    testA    5  4
## 2  Hi    testB    5  4
## 3  Lo    testA    6  4
## 4  Lo    testB    6  4
## 5  Mi    testA    7  4
## 6  Mi    testB    7  4
## 7  No    testA    8  4
## 8  No    testB    8  4

The `splits` option with `lapply` and `ddply` give me a new insight, thanks @Jake Burkhead. — Adam, Feb 27 '14 at 12:01

Descriptive Statistics by Group for multiple variables

2 Answers2

Linked