Multiple boxplots for a single plot

I have 8 matrices of nX5 dimensions that I wish to plot as boxplots on the same figure. Ideally the figure will show 8 boxplots for each of the 5 columns (i.e. 40 individual box plots). Ive struggled with trying to get the hold on functionality to work, likewise using a positional variable for each plot. Likewise Im not sure how to 'group' the individual columns from the different matrices.

PS Have attached the variable 'xvar.m' which I wish to boxplot.

Answer 1

[Kelly Kearney]

I wrote boxplot2 to handle data like this; it plots boxplots similar to grouped bar plots:

 % Some data, 8 x 1 cell array with 100 x 5 arrays

 x = cell(8,1);
 for ix = 1:8
     x{ix} = randn(100,5); 
 end

 % Plot

 boxplot2(permute(cat(3, x{:}), [2 3 1]), 1:5)

[Hillaryfor2016]

Hi Kelly,

Thank you so much for the reply. However Im having a slight bit of bother with your code and your continued assistance would be very much appreciated! I set the cell array values to my arrays that I am using

    x = cell(8,1);
x{1,1}=Fr; %11x5
x{2,1}=Kh; %1x5
x{3,1}=Me; %43x5
x{4,1}=Me2; %14x5
x{5,1}=Mi; %32x5
x{6,1}=La; %2x5
x{7,1}=Bo; %1x5
x{8,1}=Mu; %7x5
  end

Using boxplot2(permute(cat(3, x{:}), [2 3 1]), 1:5) results in the data being unsuccessfully concatenated as I would have thought given the different sizes of matrices.

Cheers and thanks once again for the reply. Ive spent my afternoon slowly questioning if I really cannot make do with a scatter plot!

PS Have attached 'xvar.m' as per star strider's request.

[Star Strider]

It may help if you save your ‘x’ variable as a .mat file and attach it (use the ‘paperclip’ (or ‘staple’) icon to upload it here, preferably to your original Question or in a Comment to it.

[Kelly Kearney]

The boxplot command ignores NaNs in its calculations, so for uneven data I usually just pad the array out with NaNs. You can do that manually or with this little function:

%%%%% Save as catuneven.m  %%%%%

function b = catuneven(dim, padval, varargin);

ndim = max(cellfun(@ndims, varargin));
ndim = max(ndim, dim);

for ii = 1:ndim
    sz(:,ii) = cellfun(@(x) size(x, ii), varargin);
end
maxsz = max(sz, [], 1);

nv = length(varargin);
val = cell(size(varargin));
for ii = 1:nv
    sztmp = maxsz;
    sztmp(dim) = sz(ii,dim);

      idx = cell(ndim,1);
      [idx{:}] = ind2sub(sz(ii,:), 1:numel(varargin{ii}));

      idxnew = sub2ind(sztmp, idx{:});

      val{ii} = ones(sztmp) * padval;
      val{ii}(idxnew) = varargin{ii};

end

b = cat(dim, val{:});

%%%%%

Then same commands as above:

load('xvar.mat');

x = catuneven(3, NaN, x{:});
boxplot2(permute(x, [3 2 1]), 1:8);

[Hillaryfor2016]

Yep, I figured this out (manual fudging) but thanks for the padding function, very helpful for future!!

Any tips on how to avoid manually colouring each individual group in the editor?

[Kelly Kearney]

The output variable returns all the handles to the objects. You can change color, linestyle, marker type, etc. using those (and I often concatenate things if I want to change a lot of them):

h = boxplot2(permute(x, [3 2 1]), 1:8);

cmap = jet(8);
tmp = struct2cell(h);
tmp = cat(3,tmp{:});
for ii = 1:8
    set(tmp(:,ii,:), 'color', cmap(ii,:));
    set(h.out(:,ii), 'markeredgecolor', cmap(ii,:));
end

[Hillaryfor2016]

Many Many thanks!!!!

[檮杌]

I love your work ! Thanks a lot!

Answer 2

[Haifeng Lu]

Thanks a lot! It really helps me!