How best to overlay bar graphs from histcounts of different data?

I wish to overlay bar graphs from two different samples that use histcounts. The length of the separate counts are not equal nor neccessarily is their bin sizes. I am using the Freedman-Diaconis bin method to ensure the best bin sizing automatically. Ultiamtely I want to best visualize the different data by color and having all the bars of the same width--and importantly in an interlaced (or side by side) fashion. I am not sure if this is possible and cannot get around doing this. I should add, I am centering the bins between the bin edges for the data and adjusted the bars to have the same width. Here's an example:

figure;
[v1, e1] = histcounts([2 5 3.4 5.3 5 4 8 7],'BinMethod','fd');
de = diff(e1)/2; % difference btwn bins
wcl1 = e1(1:end-1) + de; % bin center location
b1 = bar(wcl1,v1,0.25);
hold on;
[v2, e2] = histcounts([3.4 5.5 6.4 6.9 7],'BinMethod','fd');
de = diff(e2)/2;
wcl2 = e2(1:end-1) + de;
b2 = bar(wcl2,v2,0.25);
hold off

As it turned out the centered bin locations are identical for both data set and this visualization gives the false impression any one bar are is highlighting a breakdown of composing parts. How best to deal with such a case? Can I get the bar function to have the blue and red bars of say at column 5 side by side? In centering between the bin edges, I believe this is the best approach but I am open to any suggestion for best visualizing. Thank you in advance for any help or suggestions.

Answer 1

[Matt J]

A solution with interlacing:

[v1, e1] = histcounts([2 5 3.4 5.3 5 4 8 7],linspace(3,6,4));
[v2, e2] = histcounts([3.4 5.5 6.4 6.9 7],'BinMethod','fd');

fcn=@(z)conv(z,[1,1]/2,'valid');
e1=fcn(e1);
e2=fcn(e2);

E=unique([e1,e2]);
V1=interp1(e1,v1,E);
V2=interp1(e2,v2,E);

bar(E,[V1;V2])
xticks(E)

[hxen]

Nice! :)

I also found this that I will leave the link in case anyone encounters this and wants to read more about interleaving:

https://www.mathworks.com/matlabcentral/answers/725247-how-do-i-combine-two-bar-graphs

Answer 2

[Matt J]

Maybe make the bars semi-transparent? You can also play with the edge thicknesses to emphasize the area of inclusivity of the bars.

[v1, e1] = histcounts([2 5 3.4 5.3 5 4 8 7],'BinMethod','fd');
[v2, e2] = histcounts([3.4 5.5 6.4 6.9 7],'BinMethod','fd');

fcn=@(z)conv(z,[1,1]/2,'valid');

b1 = bar(fcn(e1),v1,'w','FaceAlpha',0.5,'LineWidth',2,'EdgeColor','r'); hold on;
b2 = bar(fcn(e2),v2,'b','FaceAlpha',0.5,'EdgeColor','none'); hold off;

[hxen]

Hi. Matt. I ended up doing that for now. But I would be shocked there isn't a work around to do the kind of interleaved plotting between two groups. If you are familiar with the stats program Prism, that is a feature built in but you have to copy and past large data sets, which is prone to copy and past errors. So working to have it as much in my analysis routines before having to use an outside program. Thank you for posting.

Answer 3

[hxen]

Thanks Matt. :)