R: Combining lattice dotplots and ordering data

Question

To create my dotplot I am using the following text file:

 ## filename  difference    RMSD
    1bso.pdb         1.0  0.5645
 1cj51.9.pdb         2.0  3.5596
 1cj51.1.pdb         3.0  3.5573
    3qzj.pdb         3.0  0.8302
    1bsy.pdb         4.0  0.5387
 1cj51.5.pdb         8.0  3.9864
    2gj5.pdb        10.0  0.8446
1cj51.10.pdb        11.0  3.5914
    1uz2.pdb        12.0  1.7741
    2blg.pdb        12.0  0.5449

The 1st column is the file name, second column in the difference and the 3rd is the RMSD. The data was ordered so the difference is ascending.

I can create individual dot plots using the following commands:

# This plots the difference
library(lattice)
data <- read.table("~/Documents/Beta_test_area/pa.txt", header=F, sep="\t")
dotplot(V1~V2, xlim=c(0, 150), xlab="CCS Difference", data=data)

# This plots the RMSD
dotplot(V1~V3, xlim=c(0, 5), xlab="RMSD", data=data)

On the graph the data on the Y axis is ordered by file name and the data is not plotted as in the text file, how can I order the Y axis to mirror the order in the data file?

The other problem I am having is combining the plots. How can I have the make the plots so that I have the plots in one row but over two columns. With the difference plot on the left and the RMSD plot on the right.

Answer 1

@Roman's part #1 is correct -- here's a slightly slick way to get the order the way you want it.

dat <- read.table(textConnection("
filename    diff RMSD
1bso.pdb    1.0 0.5645
1cj51.9.pdb 2.0 3.5596
1cj51.1.pdb 3.0 3.5573
3qzj.pdb    3.0 0.8302
1bsy.pdb    4.0 0.5387
1cj51.5.pdb 8.0 3.9864
2gj5.pdb    10.0    0.8446
1cj51.10.pdb    11.0    3.5914
1uz2.pdb    12.0    1.7741
2blg.pdb    12.0    0.5449"),
           header=TRUE)
dat <- transform(dat,filename=factor(as.character(filename),
                       levels=filename))

The grid.arrange function from the gridExtra package is handy for arranging lattice plots:

library(lattice)
d1 <- dotplot(filename~diff, xlim=c(0, 150), xlab="CCS Difference", data=dat)

# This plots the RMSD
d2 <- dotplot(filename~RMSD, xlim=c(0, 5), xlab="RMSD", data=dat)

library(gridExtra)
grid.arrange(d1,d2,nrow=1)

Or (from @Aaron):

library(latticeExtra)
c(d1,d2)

Alternatively as @Roman suggested you can create small multiples.

library(reshape)
m <- melt(dat)
dotplot(filename~value|variable,
       scales=list(x=list(relation="free")), xlim=list(c(0,150), c(0,5)),
       data=m)

Or

library(ggplot2)
g1 <- qplot(value,filename,data=m)+
  facet_grid(.~variable,scale="free")+theme_bw()+
  opts(panel.margin=unit(0,"lines"))

although here I really don't know how to set the x axis limits panel-by-panel, other than doing something nasty like trying to add invisible points appropriately.

edit: panel-by-panel scaling from Josh O'Brien, latticeExtra from Aaron

Answer 2

I think your first question is related to ordering of factors. It's a common problem but once you learn the trick that factors use, it become a (nice) feature. This has been discussed a number of times, at least here and here.

I'm not sure I understand your second question down to all the details, but generally there are two strategies. In base graphics, you can use par argumentmfrow to open a device with defined rows/columns into which you plot your graphics, e.g. par(mfrow = c(2, 1)) which will plot two plots in two rows and one column. par(mfrow = c(2,2)) will give you graphs laid out in 2x2 grid. You can also consider alternatives, layout and split.screen.

In grid graphics (think lattice and ggplot2), the approach is different. You can plot a number of graphs in a grid, using | or facet_grid for lattice and ggplot2, respectively.