Boxplot groups by mask set

I have a vector of data, and a set of masks defining my (possibly overlapping) groups:

data = rand(100,1)
grpMasks = [(1:100)'<25,  (1:100)'<60&(1:100)'>10,  (1:100)'>40,  rand(100,1)>0.5]

So my data is 100-by-1, I have 4 groups in a 100-by-4 logical mask. I want to make a boxplot with 4 boxes, 1 per mask.

boxplot(data, grpMasks)

Makes N boxes, where N is the number of unique row combinations in grpMasks.

I know that I can loop from 1 to 4, make a single boxplot and then shift each box's X-location, but I feel like that's unnecessarily messy. I've got my data, I've got a format of my groups. Can boxplot handle this form of grouping?

Answer 1

[Sven]

Ah! Here's my solution. Repeat the data for as many groups as you have, and then NaN out the data not belonging to each group:

data = rand(100,1);
grpMasks = [(1:100)'<25,  (1:100)'<60&(1:100)'>10,  (1:100)'>40,  rand(100,1)>0.5];
numGrps = size(grpMasks,2);
allData = data * ones(1,numGrps);
allData(~grpMasks) = nan;
boxplot(allData);