tweaking colours and scales of heatmap in R

Question

I am making an heatmap using R and the function levelplot() from lattice.

The commands I am using are:

data<-read.table("data_corr", header=T) 
library(gplots)
library(lattice)
z<-cor(data)
levelplot(z)

and I get a figure like this

Three things:

(a) Row labels are smudged, how this can be resolved to get more clear row labels?

(b) How can I use the colours red, green and blue instead of default (pink, white and blue)?

(c) Is it possible to tweak the scale from 0.90 to 1 instead of 0.84?

Solutions in python are also welcomed, in that case I have a correlation file of 90*90 (row * column), say filename is z.txt.

Answer 1

In python's matplotlib, the answers to your questions are:

(a) Plot with the imshow option interpolation="nearest"

(b) Colormaps can be chosen in the imshow option cmap=cm.jet for example. More examples here.

(c) To scale, I'm assuming you want to only show values in the threshold and have the colorbar map accordingly. Use imshow option vmax and vmin.

Here is a minimal working example:

import numpy as np
import pylab as plt

# Create random matrix, save it to a text file
N = 90
A = np.random.random((N,N))
np.fill_diagonal(A,.95)
np.savetxt("z.txt",A)

# Load the data
A = np.loadtxt("z.txt")

# Apply a mask so values below .9 don't show
A[A<.9] = None

# Scale the values from .9 to 1.0 using vmin,vmax
plt.imshow(A, interpolation='nearest', vmin=.9,vmax=1.0,cmap=plt.cm.jet)
plt.colorbar()
plt.show()

Answer 2

From the help page for levelplot:

col.regions
color vector to be used if regions is TRUE. The general idea is that this should be a color vector of moderately large length (longer than the number of regions. By default this is 100). It is expected that this vector would be gradually varying in color (so that nearby colors would be similar). When the colors are actually chosen, they are chosen to be equally spaced along this vector. When there are more regions than colors in col.regions, the colors are recycled. The actual color assignment is performed by level.colors, which is documented separately.

Answer 3

In addition to Carl's answer:

• you can set the default for col.regions by e.g.

  `trellis.par.set (regions = list (col = rainbow))
  

• range of the colour scale: see levelplot's parameter at.

• row and column names: see ? xyplot, parameter scales.