Color Different Scatter Points Based on Group

Hello,

I have a great deal of points on a 3D scatter plot that I need to be shown in different colors. They need to be categorized in color based on a label that they are given in the data set. All examples I have seen for coloring markers in scatter plots has been based on distance/position (which is not what I need).

Any help with how to code color markers based on grouping/labeling would be appreciated!

Answer 1

[Walter Roberson]

[unique_groups, ~, group_idx] = unique(Your_Group_Labels);
num_groups = size(unique_groups, 1);

if isnumeric(unique_groups)
   group_names = cellstr( num2str( unique_groups ) );
elseif ischar(unique_groups)
  group_names = mat2cell( unique_groups, ones(1, num_groups), size(unique_groups, 2) );
elseif iscell(unique_groups)
  group_names = unique_groups;
else
  error('I do not know how to convert group labels of class "%s" to printable strings', class(unique_groups));
end

pointsize = 30;
scatter3(x, y, z, pointsize, group_idx);

cmap = jet(num_groups);    %or build a custom color map
colormap( cmap );

legend(group_names);

If only you knew ahead of time what the data class of your labels was going to be, then you could simplify your code...

[Brian Tomblin]

I am unsure what you mean exactly with data class of the labels, but they are text names of muscles in a cell array, if that helps. The idea is to color-separate the XYZ coordinates based on their generated labeled muscle.

[Walter Roberson]

For the case of the labels being cell array of strings:

[unique_groups, ~, group_idx] = unique(Your_Group_Labels);
num_groups = size(unique_groups, 1);

pointsize = 30;
scatter3(x, y, z, pointsize, group_idx);

cmap = jet(num_groups);    %or build a custom color map
colormap( cmap );

legend(unique_groups);

[Brian Tomblin]

If I already have a legend with corresponding labels and colors produced by the original code, can that be used more easily in the scatter3?

[Walter Roberson]

If you have a cell array of (unique) labels, and an N x 3 RGB array of corresponding color specifications, then the easiest way would be to sort() the (unique) label names and pull out the sorting order (second output of sort()) and use that order to rearrange the color table rows to create the cmap variable above, and also use the order to re-arrange the legends. To phrase that a different way, you should create a colormap variable whose rows are in the same order as the alphanumeric sort order as the labels themselves, and likewise a rearranged legend

[~, sortidx] = sort(label_name_table);
ordered_cmap = corresponding_color_table(sortidx, :);
ordered_legends = corresponding_legend_entries(sortidx);
[~, ~, group_idx] = unique(Your_Group_Labels);

pointsize = 30;
scatter3(x, y, z, pointsize, group_idx);

colormap( ordered_cmap );
legend( ordered_legends );

Answer 2

[Chad Greene]

Similar to Walter's solution:

% define 25 random points: 
x = randn(25,1); 
y = randn(25,1); 

% Each x,y point is in group 1, group 2, or group 3: 
group = randi(3,25,1); 

% Plot: 
plot(x(group==1),y(group==1),'.','markersize',20,'color','green'); 
hold on
plot(x(group==2),y(group==2),'.','markersize',20,'color','red'); 
plot(x(group==3),y(group==3),'.','markersize',20,'color','black'); 
legend('group 1','group 2','group 3')

[Walter Roberson]

If you have 2D data then you can use gscatter()

[Chad Greene]

Ah, yes, I forgot about that function!