Setting up a CSV file for R to display histograms

Question

Greetings,

Basically, I have two vectors of data (let's call it experimental and baseline). I want to use the lattice library and histogram functions of R to plot the two histograms side-by-side, just as seen at the end of this page.

I have my data in a CSV file like this:

Label1,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18
Label2,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18
Label3,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18
Label4,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18
Label5,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18
Label6,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18

Each row should be a new pair of histograms. Columns 1-9 represents the data for the experiment (left-side histogram). Columns 10-18 represents the baseline data (right-side histogram).

Can anyone help me on this? Thanks.

Answer 1

Something like this should work:

library(lattice)
data <- matrix(1:18, ncol=18, nrow=3, byrow=T)

for (i in 1:nrow(data))
    {
    tmp <- cbind(data[i,], rep(1:2, each=9))
    print(histogram(~tmp[,1]|tmp[,2]), split=c(1,i,1,nrow(data)), more=T)
    }

Note: this will work only for few rows of data... for larger datasets you may want to think of slightly different layout (change the split parameter in histogram)

Answer 2

Your data is poorly formatted for faceting with lattice. You can restructure it using reshape.

read.csv(textConnection("Label1,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18
Label2,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18
Label3,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18
Label4,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18
Label5,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18
Label6,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18"), header = F)->data

colnames(data)[1] <- "ID"
colnames(data)[2:10] <- paste("exp",1:9, sep = "_")
colnames(data)[11:19] <- paste("base", 1:9, sep = "_")

library(reshape)
data.m <- melt(data, id = "ID")
data.m <- cbind(data.m, colsplit(data.m$variable, "_", names = c("Source","Measure")))

data.m is now in the format you really want your data to be in for almost everything. I don't know if each of the 9 measurements from the experiment and the baseline are meaningful or can be meaningfully compared so I kept them distinct.

Now, you can use lattice properly.

histogram(~value | Source + ID, data = data.m)

If the measurements are meaningfully compared (that is, data[,2] and data[,11] are somehow the "same"), you could recast the data to directly compare experiment to baseline

data.comp <- cast(data.m, ID + Measure ~ Source)  
## I know ggplot2 better
library(ggplot2)
qplot(base, exp, data = data.comp)+
  geom_abline()+
  expand_limits(x = 0, y = 0)