Grouped boxplots in Matlab: a Generic function - matlab

Question

After seeing this great post in SO:
Most efficient way of drawing grouped boxplot matlab
I was wondering if it is possible to create a function like that but a bit more generic, as in my application I need to make several analysis of different algorithms in different situations and it would be very tedious to tune the plotting code for each case.
I would like something generic for this kind of plots:

Answer 1

I coded a Matlab function that does that for you (me).
Features:
In each boxplot different amount of data supported
Any amount of groups and boxplot per group supported
Xlabel and boxplotlabel supported
Automatic choice of colors or user specified colors
Example of result of function:
CODE:
function multiple_boxplot(data,xlab,Mlab,colors)
% data is a cell matrix of MxL where in each element there is a array of N
% length. M is how many data for the same group, L, how many groups.
%
% Optional:
% xlab is a cell array of strings of length L with the names of each
% group
%
% Mlab is a cell array of strings of length M
%
% colors is a Mx4 matrix with normalized RGBA colors for each M.
% check that data is ok.
if ~iscell(data)
error('Input data is not even a cell array!');
end
% Get sizes
M=size(data,2);
L=size(data,1);
if nargin>=4
if size(colors,2)~=M
error('Wrong amount of colors!');
end
end
if nargin>=2
if length(xlab)~=L
error('Wrong amount of X labels given');
end
end
% Calculate the positions of the boxes
positions=1:0.25:M*L*0.25+1+0.25*L;
positions(1:M+1:end)=[];
% Extract data and label it in the group correctly
x=[];
group=[];
for ii=1:L
for jj=1:M
aux=data{ii,jj};
x=vertcat(x,aux(:));
group=vertcat(group,ones(size(aux(:)))*jj+(ii-1)*M);
end
end
% Plot it
boxplot(x,group, 'positions', positions);
% Set the Xlabels
aux=reshape(positions,M,[]);
labelpos = sum(aux,1)./M;
set(gca,'xtick',labelpos)
if nargin>=2
set(gca,'xticklabel',xlab);
else
idx=1:L;
set(gca,'xticklabel',strsplit(num2str(idx),' '));
end
% Get some colors
if nargin>=4
cmap=colors;
else
cmap = hsv(M);
cmap=vertcat(cmap,ones(1,M)*0.5);
end
color=repmat(cmap, 1, L);
% Apply colors
h = findobj(gca,'Tag','Box');
for jj=1:length(h)
patch(get(h(jj),'XData'),get(h(jj),'YData'),color(1:3,jj)','FaceAlpha',color(4,jj));
end
if nargin>=3
legend(fliplr(Mlab));
end
end
Simple example:
clear;clc;
% Create example data
A=rand(100,10);
B=rand(200,10);
C=rand(150,10);
% prepare data
data=cell(10,3);
for ii=1:size(data,1)
Ac{ii}=A(:,ii);
Bc{ii}=B(:,ii);
Cc{ii}=C(:,ii);
end
data=vertcat(Ac,Bc,Cc);
xlab={'Hey','this','works','pretty','nicely.','And','it','has','colors','!!!!'};
col=[102,255,255, 200;
51,153,255, 200;
0, 0, 255, 200];
col=col/255;
multiple_boxplot(data',xlab,{'A', 'B', 'C'},col')
title('Here it is!')
Mathworks file exchange file can be found here:
http://www.mathworks.com/matlabcentral/fileexchange/47233-multiple-boxplot-m