Create a grouped boxplot

I have 3 groups of data A, B, and C.

A = rand(100,8);
B = rand(100,8)*2;
C = rand(100,8)*4;

I want to plot Box chart of grpup A,B and C.

The internal label of each group is {'a' 'b' 'c' 'd' 'e' 'f' 'g' 'h'}.

I found boxplotgroup(), and used below code

data = {A,B,C};
boxplotGroup(data, 'PrimaryLabels', {'a' 'b' 'c' 'd' 'e' 'f' 'g' 'h'}, ...
  'SecondaryLabels',{'A', 'B' 'C'}, 'InterGroupSpace', 3)

But I got this error.

Error using assert (line 5)
The number of primary labels must equal either the number of bars per group
(3) or the number of total bars (24).

Error in boxplotGroup (line 155)
    assert(ismember(numel(p.Results.primarylabels),[nMembers,
    nMembers*nGroups]), ...

Error in (line 13)
boxplotGroup(data, 'PrimaryLabels', {'a' 'b' 'c' 'd' 'e' 'f' 'g' 'h'}, ...
 

Answer 1

[Christopher McCausland]

Hi NA,

From the documentation here (See section boxplotGroup(__,'primaryLabels',c) ) the number of primary lables should match the number of boxplotGroups. i.e. you appear to be plotting three box plots (A,B,C) but trying to give these box plots eight names ('a' 'b' 'c' 'd' 'e' 'f' 'g' 'h').

I beleive that you are trying to show something like the last example in boxplotGroup(__,'secondaryLabels',s) (shown below) in which case swap your primary and secondary labels.

Let me know if that helps,

Christopher

[NA]

Actually, I want this plot. (Suppose the blue bands are box plot)

[Adam Danz]

@Christopher McCausland thanks for providing support.

@NA I know the boxplotGroup inputs are not intuitive. I may try to improve that in the future while maintaining backward compatibility.

If A B and C are all the same size, here's the solution. If they aren't the same size, you need to pad the smaller matrices with NaN values to make the equal.

A = rand(100,8);
B = rand(100,8)*2;
C = rand(100,8)*4;

D = squeeze(mat2cell(permute(cat(3,A,B,C),[1,3,2]),size(A,1),3,ones(1,size(A,2))));
%                                        Number of matricies ^

clf
boxplotGroup(D','PrimaryLabels', {'a' 'b' 'c' 'd' 'e' 'f' 'g' 'h'}, ...
  'SecondaryLabels',{'A', 'B' 'C'}, 'InterGroupSpace', 3)

[Edward Jero Sam Jeeva Raj]

Hi, The subscript command used in primary lables are not working. Ex, i gave the lables like 'A_{a}' where a is the subscript of A, but it is printed A_{a} in boxplot label. Could you pls help in this. Thanks, Have a nice day

[Christopher McCausland]

Hi Edward,

This question should really be asked as a question in it's own right and not as an answer to a previously solved question.

I would however suggest making the labels and supplying them to the primary labels input rather than trying to make them on the fly.

Here is some further help with a previously solved similar problem.

Kind Regards,

Christopher

[Adam Danz]

@Edward Jero Sam Jeeva Raj

To fix this, set the axis tick label interpreter to tex

h = boxplotGroup(__);

h.axis.TickLabelInterpreter = 'tex';

Answer 2

[Baldvin]

There are a few ways to accomplish this (or something similar looking) by using boxchart itself. Let's start with the same dummy data as in the original question:

nRows=100;
nCols=8;
nMat=3;

A = rand(nRows,nCols);
B = rand(nRows,nCols)*2;
C = rand(nRows,nCols)*4;
% Stack the data into a single vector:
dataAsVector = [A(:);B(:);C(:)];

% Modify these if the dimensions are changed:
columnLabels = ["a","b","c","d","e","f","g","h"];
matrixLabels = ["A","B","C"];

In the examples below, we'll be "chopping up" the data in dataAsVector, so it might help to have a visual of what it contains. The columns of each matrix are stacked on top of each other, that is, dataAsVector(1:nRows) is the first column of A, and so forth.

Furthermore, it might be helpful to retrace the steps with smaller matrices (i.e. just use two matrices each with two columns) and inspect the inputs to boxchart. In that case, we could use different labels:

columnLabels = strcat("c",string(1:nCols));

matrixLabels = strcat("M",string(1:nMat));

Let's start with creating a single boxchart, where we manually place the columns at locations 1, 2, 3, etc. (of which there are nCols*nMat = 24), using the xgroupdata input:

bSimple = boxchart(categorical(repelem(1:(nCols*nMat),nRows)),dataAsVector);
% Change the labels:
ax = gca;
ax.XTickLabel = repmat(columnLabels,1,nMat);

% Optional: fiddle with manually placing sublabels. This will have to be
% modified if we change the number of columns or number of matrices
ax.XTickLabel([4 12 20]) = {'d\newlineA','d\newlineB','d\newlineC'};

Next, we'll use xgroupdata to distinguish the columns, and GroupByColor to distinguish the matrices. Note that ColorGroupLayout came out in R2025a. The plot below contains three boxchart objects, one per matrix:

figure
b3 = boxchart(categorical(repelem(1:(nCols*nMat),nRows)),dataAsVector,...
    'GroupByColor',repelem(matrixLabels,nRows*nCols),'ColorGroupLayout','overlaid');
legend('Location','northeastoutside')
ax = gca;
ax.XTickLabel = repmat(columnLabels,1,nMat);

Next, we'll use xgroupdata to distinguish the matrices, and GroupByColor to distinguish the columns. This creates nCols boxchart objects, i.e. one for each column. Setting the SeriesIndex makes them all have the same color.

figure
b8 = boxchart(repelem(categorical(matrixLabels),nRows*nCols),dataAsVector, ...
    'GroupByColor',repmat(repelem(columnLabels,nRows),1,nMat),...
    'SeriesIndex',1);

I hope this helps a little bit using boxchart.