R - d3heatmap - implement breaks - r

Question

I am trying to plot a heatmap using the d3heatmap package.
Unfortunately, I have not been successful yet in implementing certain breaks using the option breaks=... as in heatmap or heatmap.2.
This yields just funny results, I am not even sure whether I am doing something wrong or whether the function just ignores breaks.
For example, I tried:
breaks = c(seq(-10, -2), seq(-2, -1.65), seq(-1.65, 1.65), seq(1.65, 2), seq(2, 10)
and
breaks = c(-10, -2, -1.65, 1.65, 2, 10)
with
colors = c("red", "yellow", "green", "yellow", "red")
but nothing seems to work properly.
Any suggestions?
Here's the dput of my data:
> dput(mat)
structure(c(-0.04, NA, 0.59, NA, 0.675, 0.96, 1.09, 0.445, NA,
0.545, NA, NA, 0.09, -1.11, NA, 0.99, 0.13, 0.215, 1.425, 0,
NA, 0.69, 0.805, NA, 0.69, 1.22, NA, 0.3, NA, 0.025, NA, 0.075,
0.36, -0.94, NA, -0.31, 0.26, 1.02, -1.19, NA, NA, -0.77, NA,
-1.48, 1.05, 0.48, NA, NA, NA, 1.49, -1.285, NA, 0.76, 1.14,
-0.62, NA, NA, NA, 0.95, NA, NA, -0.12, 0.49, NA, 2.31, NA, -0.33,
0.85, NA, -1.7, -1.63, NA, -1.12, 0.135, -0.18, NA, -0.245, NA,
-0.2, -0.2, 0.23, -0.11, NA, 0.3, -0.81, 0.04, 0.18, -0.7, 0.53,
0.44, -0.49, 0.28, 0.26, 0.06, 0.265, 0.21, 0.06, -0.175, 0.365,
0.255, 1.25, -0.35, 0.16, 0.125, 0.825, 0.08, 0.02, -0.02, 0.99,
0.79, -0.23, 0.06, NA, 0.36, -0.64, -0.195, 1.19, -0.29, 0.915,
NA, NA, NA, NA, 0.2, 0.1, NA, 0.04, 0.33, NA, 1.46, 2.36, NA,
-0.92, 1.295, NA, NA, 0.8, NA, 1.09, 1.45, 5.42, NA, NA, NA,
1.69, 3.43, NA, 0.55), .Dim = c(37L, 4L), .Dimnames = list(c("AT",
"BE", "BG", "CEE", "CH", "CN", "CZ", "DE", "DK", "EA", "EE",
"EMU", "ES", "EU", "FI", "FR", "GB", "GR", "HR", "HU", "IE",
"IT", "JP", "LU", "NL", "PL", "PT", "RO", "RS", "RU", "SE", "SI",
"SK", "TR", "UA", "UK", "US"), c("Credit Risk", "Funding and liquidity Risk",
"Macro Risk", "Market Risk")))
And the code I am running:
d3heatmap(abs(mat),
dendrogram = "none",
breaks = c(0,1.65,2,10),
col = c("green", "yellow", "red"),
na.rm = TRUE)
The same function using heatmap.2 works perfectly, though.

Answer 1

The function d3heatmap simply does not have a 'breaks' argument. If it gets passed in as an argument it is silently ignored. (See ?d3heatmap.)
The heatmap.2 function in the gplots package on the other hand does have a "breaks" argument. That explains the difference in behaviour.
Luckily, it is still possible to get the desired behaviour by passing an appropriate 'colors' function to d3heatmap. It works as follows.
First the example data:
mat <- structure(c(-0.04, NA, 0.59, NA, 0.675, 0.96, 1.09, 0.445, NA,
0.545, NA, NA, 0.09, -1.11, NA, 0.99, 0.13, 0.215, 1.425, 0,
NA, 0.69, 0.805, NA, 0.69, 1.22, NA, 0.3, NA, 0.025, NA, 0.075,
0.36, -0.94, NA, -0.31, 0.26, 1.02, -1.19, NA, NA, -0.77, NA,
-1.48, 1.05, 0.48, NA, NA, NA, 1.49, -1.285, NA, 0.76, 1.14,
-0.62, NA, NA, NA, 0.95, NA, NA, -0.12, 0.49, NA, 2.31, NA, -0.33,
0.85, NA, -1.7, -1.63, NA, -1.12, 0.135, -0.18, NA, -0.245, NA,
-0.2, -0.2, 0.23, -0.11, NA, 0.3, -0.81, 0.04, 0.18, -0.7, 0.53,
0.44, -0.49, 0.28, 0.26, 0.06, 0.265, 0.21, 0.06, -0.175, 0.365,
0.255, 1.25, -0.35, 0.16, 0.125, 0.825, 0.08, 0.02, -0.02, 0.99,
0.79, -0.23, 0.06, NA, 0.36, -0.64, -0.195, 1.19, -0.29, 0.915,
NA, NA, NA, NA, 0.2, 0.1, NA, 0.04, 0.33, NA, 1.46, 2.36, NA,
-0.92, 1.295, NA, NA, 0.8, NA, 1.09, 1.45, 5.42, NA, NA, NA,
1.69, 3.43, NA, 0.55), .Dim = c(37L, 4L),
.Dimnames = list(c("AT", "BE", "BG", "CEE", "CH", "CN", "CZ", "DE", "DK", "EA", "EE", "EMU", "ES", "EU", "FI", "FR", "GB", "GR", "HR", "HU", "IE", "IT", "JP", "LU", "NL", "PL", "PT", "RO", "RS", "RU", "SE", "SI", "SK", "TR", "UA", "UK", "US"), c("Credit Risk", "Funding and liquidity Risk", "Macro Risk", "Market Risk")))
Suppose we want the following three color bins: blue for values < 0, green for values >= 0 but < 2, and red for values >= 2. We then define the corresponding ordered list of colors.
palette <- c("blue", "green", "red")
We also define the boundary values of the color bins. These values must include the domain boundaries.
mi <- min(mat, na.rm = TRUE)
ma <- max(mat, na.rm = TRUE)
breaks <- c(mi, 0, 2, ma)
We can now define a color interpolation function which maps a value in [0,1] onto a color, respecting our color bins. The 'scales' package comes to help here.
install.package('scales') # if needed
library(scales)
colorFunc <- col_bin(palette, bins = rescale(breaks))
The breaks originally defined in the domain of our data needed to be rescaled to [0,1]. The 'rescale' function in the 'scales' package handled that.
Small detail: the low boundary of a bin is included in the bin, but the high boundary is excluded. So the value 0 will be green, anything between 0 and 2 will be green too, but 2 will be red.
We can now plot the heat map.
d3heatmap(mat, dendrogram = "none", colors = colorFunc, na.rm = TRUE)
The result looks like this: