Question

I have data frame that looks like this:

set.seed(123)
df <- data.frame(factor1 = rep(c("A", "B"),50),
                 factor2 = rep(c("X","X", "Y", "Y"),25),
                 value = rnorm(100))

I want to calculate some summary values for factor1:factor2 pairs. I calculated the means and sd using:

summary <- as.matrix(cast(df, factor1~factor2, mean))
summary.sd <- as.matrix(cast(df, factor1~factor2, sd))
summary.table <- t(rbind(summary, summary.sd))
colnames(summary.table) <- c("A.mean", "B.mean", "A.sd", "B.sd")

But I would like to add to summary.table the p-value from a t.test comparing A vs B. So far I have done this, but not only does this not write to summary.table but i can’t get the names of the factor2 variable to print out along with it:

for (measurement in levels(df$factor2)) print(t.test(value~factor1, data=subset(df, factor2==measurement)))

I figure there must be some simple way to do this, or maybe a package that I am not aware of that would make this much more straightforward.

Answer 1

I’d do it this way:

First, get mean and sd summaries using ddply from plyr using summarise

require(plyr)
require(reshape2)
o1 <- ddply(df, .(factor1, factor2), summarise, mean = mean(value), sd=sd(value))

#   factor1 factor2       mean        sd
# 1       A       X 0.03746854 0.8730525
# 2       A       Y 0.18352432 0.7635439
# 3       B       X 0.10317706 1.0494930
# 4       B       Y 0.03745372 0.9876173

Then, get p-values for t-test with NULL mean(A) = mean(B) for both X and Y levels in factor2:

o2 <- ddply(df, .(factor2), summarise, pval=t.test(value ~ factor1)$p.value)

#   factor2      pval
# 1       X 0.8108754
# 2       Y 0.5614256

Then, using reshape2‘s melt and dcast cast o1 to the desired format.

o1.mc <- dcast(melt(o1, c("factor1", "factor2")), factor2 ~ variable + factor1)

#   factor2     mean_A     mean_B      sd_A      sd_B
# 1       X 0.03746854 0.10317706 0.8730525 1.0494930
# 2       Y 0.18352432 0.03745372 0.7635439 0.9876173

Now, merge it with o2:

merge(o1.mc, o2)

#   factor2     mean_A     mean_B      sd_A      sd_B      pval
# 1       X 0.03746854 0.10317706 0.8730525 1.0494930 0.8108754
# 2       Y 0.18352432 0.03745372 0.7635439 0.9876173 0.5614256