transform bar chart with separate bars (for each value of y) to bar chart with overlapping bars

Hi. I need to transform this figure:

load matrix.mat
matrix = matrix(5:7,:);

figure();
cm  = [1,0,0; 0,0,1; 0,1,0; 1,0,0.52];
hbh = barh(matrix(:,1), matrix(:,[2 3 4 5]));
for k =  1:numel(hbh)
    hbh(k).FaceColor = cm(k,:);
    hbh(k).EdgeColor = [0,0,0];
end

into this one:

I tried the following way but I don't get the same result.

figure
hbh = barh(matrix(:,1), matrix(:,[2 3 4 5]),'stacked');
cm  = [[1,0,0]; [0,0,1]; [0,1,0]; [1,0,0.52]];
for k =  1:numel(hbh)
    hbh(k).FaceColor = cm(k,:);
    hbh(k).EdgeColor = [0,0,0];
end

[Dyuman Joshi]

If I understand it correctly, you want to

- have gaps between stacked plots

- reorder the graphs in the order > green, blue, red and the custom color from left to right

- delete the green part from the top most bar

Is that correct?

[Alberto Acri]

Hi @Dyuman Joshi

No for all three answers.

I want to overlap the 4 bars for each value on the y-axis. Overlap the 4 bars at the value 66, overlap the 4 bars at the value 67 and overlap the 4 bars at the value 68.

The second figure is indicative but it is close to the result I would like to get.

The third figure is what I would like to achieve but the values on the x-axis are not correct.

[Rik]

So what you actually want to stack is equivalent to this?

mag = 250;
gre = 50;
blu = 45;
red = 170;
data = [gre blu red mag];
DataToPlot = max(0,[data(1) diff(data)])

DataToPlot = 1×4

50 0 125 80

There seems to be something wrong with the order of the segments, since the rest of your plot suggest blue is more to the front than green.

You can also achieve this by using hold and plotting in the order you want. You might not end up with the exact visuals you want, but you can try to see whether it is something you would like.

[the cyclist]

I interpreted the request as

• colors are associated with the original column location
• plot each row in sorted value order (shifting color accordingly, as needed)

If this is true, then there is a sliver of green for the row at value 68.

In your example, @Rik, it would be

mag = 250;
gre = 50;
blu = 45;
red = 170;
data = sort([gre blu red mag]);
DataToPlot = [data(1) diff(data)] % Don't need max(), because there are no negative diffs

DataToPlot = 1×4

45 5 120 80

The color order would be [blu, gre, red, mag].

That's the algorithm I implemented in my solution.

[Rik]

The reason I included the max call was for robustness. If at one point a negative diff occurs, I would like to prevent unexpected behavior.

[the cyclist]

I understand. But after the sort, a negative diff can't occur. (But if I am wrong about the sorting, then the max would be required, as you state.)

Answer 1

[the cyclist]

load matrix

M = matrix(:,2:end);

[sortedM,sortingIndex] = sort(M,2);

cm  = [[1,0,0]; [0,0,1]; [0,1,0]; [1,0,0.52]];
figure
hold on
for mi =  1:height(sortedM)
    hbh = barh(matrix(mi,1), [sortedM(mi,1),diff(sortedM(mi,:),1,2)],'stacked');
    for k = 1:width(sortedM)
        hbh(k).FaceColor = cm(sortingIndex(mi,k),:);
    end
end

[the cyclist]

I stuck to answering what you asked for, but be aware that this method of plotting bars "in front of each other" can be somewhat deceptive, from an information point of view. For example, the green bar at 68 is nearly the same value as the blue bar, but the visual impact is tiny. (I think that's the reason @Dyuman Joshi thought you didn't want it at all.)

It would take more work, but you might want to consider using the method of overlaying bars that is described on this documentation page.

[Alberto Acri]

Thank you @the cyclist. That is the result I wanted.

Right now I prefer to use the graph with the bars overlaid with the code you provided because I have the desired values on the x-axis. Of course, it is not easy to read, in fact I will modify it with the documentation you attached.

I asked the question here because I have a problem with the visualization. If you want you can take a look at it.