Separate categories in heatmap colorscale

Hello,

I am using a lot heatmaps to illustrate matrices of data such as this one. I often run in the scenario where I could use a data value that would appear similar to NaN but in another color. Here I gave the tiles the value -10 and troncated the colorscale, but I was wondering if anyone knows of a cleaner way to do it because this method is not always optimal.

Thanks !

Answer 1

[Voss]

One way would be to put a patch object over the particular cells of the heatmap that you want to be different. Here's a way to do that:

% first, a matrix with some NaN elements:
M = magic(10);
M([1 2 3 11 12 21]) = NaN;

% create the heatmap:
h = heatmap(M);

% color to use to obscure some particular cells of the heatmap:
color = 'w';

% let's say these are the linear indices of the elements in M you 
% would replace with -10 (or whatever special value) and color with 
% the specified color (I just picked some locations in the lower-
% right of the heatmap):
idx = [70 79 80 86 89 90 95 96 97 98 99 100];

disp(idx(:));

    70
    79
    80
    86
    89
    90
    95
    96
    97
    98
    99
   100

% convert those linear indices to rows and columns:
[m,n] = size(M);
[ridx,cidx] = ind2sub([m,n],idx);

disp([ridx(:),cidx(:)]);

    10     7
     9     8
    10     8
     6     9
     9     9
    10     9
     5    10
     6    10
     7    10
     8    10
     9    10
    10    10

% you can't put a patch into a heatmap chart, so create a new invisible 
% axes on top of the heatmap, with the same size, limits, and 
% orientation (YDir) as the heatmap:
ax = axes('Units','normalized','Position',h.InnerPosition, ...
    'Visible','off','XLim',[0.5 n+0.5],'YLim',[0.5 m+0.5],'YDir','reverse');

% put a patch in the new axes, to obscure those particular cells in the
% heatmap, using the specified color:
xd = 1:n;
yd = 1:m;
patch(ax,xd(cidx)+[0;0;1;1]-0.5,yd(ridx)+[0;1;1;0]-0.5,color)