Legend/colorbar for scatterplot with colour-coded subject-wise markers

I am simplifying my problem to make it easier to answer. Say I have ratings on two measures, x and y, from N subjects. I would like to do a scatterplot of x and y with a different marker colour for each subject, and display a colorbar/legend that shows what colour corresponds to which subject. I have the following code, that I hoped would assign a random colour to each subject across a given colour-space:

N_subj = 30;
X = rand(30,20);
Y = rand(30,20);
for i_subj=1:N_subj
  x = X(i_subj,:);
  y = Y(i_subj,:);
  hold on
  markerColour = i_subj/N_subj;
  markerColour = [markerColour markerColour markerColour];
  colormap summer
  h(i_subj)=scatter (x, y, ...
    'MarkerEdgeColor','black',...
    'MarkerFaceColor',markerColour);
end
colorbar

However, this does not happen so in the code above I instead chose increasingly dark shades of grey. Problem, is the colorbar command displays an irrelevant (yellow-to-red) colour bar, instead of the shades of grey that I used.

If I use legend instead of colorbar, the colour associations are correct but the legend entries are discrete whereas I;d like the mto be continous (stacked rectangles).

Any recommendations? Many thanks

Answer 1

[Image Analyst]

When you say

markerColour = [markerColour markerColour markerColour];

R, G, and B are all the same - thus your color is some darkness of gray. Perhaps you want this:

N_subj = 30;
X = rand(30,20);
Y = rand(30,20);
% Create colormap with N_subj colors.
summerColorMap = summer(N_subj);
for i_subj=1:N_subj
  x = X(i_subj,:);
  y = Y(i_subj,:);
  hold on
  % Get the color for this subject from the summer colormap.
  markerColour = summerColorMap(i_subj,:);
  h(i_subj)=scatter (x, y, ...
    'MarkerEdgeColor','black',...
    'MarkerFaceColor',markerColour);
end
% Display colorbar.
colormap(summerColorMap);
colorbar

[z8080]

Many thanks! I had indeed realised that my markerColour variable meant the colour-coding was grey-scale; I just thought it would make the plot cleaner, but in fact it's harder to distinguish between the 30 shades, as you lack the hue dimension. Your solution uses the summer colormap, just as I wanted. Two more questions:

1) Since it's still hard to distnguish between N=30 shades, is there a colormap whose ends' hues are such as to maximise the difference between the colourmap's ends?

2) How can I have the colourmap labelled in terms of subject number (1 to 30) instead of continuously 0 to 1?

Thanks again.

[z8080]

Any thoughts on these two follow up questions? Many thanks!

[Image Analyst]

Try lines(), colorcube(), prism(), or hsv(). See the list of them in the help for colormap.

Or see this link, or make your own.

To label the colorbar, see the help for colorbar where it shows this example:

colorbar('Ticks',[-5,-2,1,4,7],...
       'TickLabels',{'Cold','Cool','Neutral','Warm','Hot'})