Two bar graphs - two data sets - different bin width

Hi everyone,

I am kinda new to matlab and may not have been asking the right question to google. Anyway, here is my problem: I've got two data sets called one and two:

none = length(one)
ntwo = length(two)
%h = nan(max(none,ntwo),2)
h(1:none,1)=one
h(1:ntwo,2)=two

From there I extract the best bin width and the theoretical best amount of bin for each data set ( following the The Freedman-Diaconis rule):

%one -------
minNumberOne = round(min(h(1:none,1)))
maxNumberOne = ceil(max(h(1:none,1)))
iqrVarOne = iqr(h(1:none,1))
hOne = ceil(2 * iqrVarOne * none^(-1/3))
binsOne = ceil((maxNumberOne -  minNumberOne)/hOne)

%two -------
minNumberTwo = round(min(h(1:ntwo,2)))
maxNumberTwo = ceil(max(h(1:ntwo,2)))
iqrVarTwo = iqr(h(1:ntwo,2))
hTwo = ceil(2 * iqrVarTwo * ntwo^(-1/3))
binsTwo = ceil((maxNumberTwo -  minNumberTwo)/hTwo)

From now I am a bit lost. I'd like to apply for each bar graph the number of bin and the bin width, AND displaying those graph on the same figure, like that:

My code:

figure(1)
data = [h(:,1) h(:,2)]
[y, x] = hist(data)
bar(x,y, 'group')

this code doesn't take into account the bins calculated before of course. Do you have any ideas on how to integrate that to my code?

Cheers everyone. I hope I have been clear enough...

Flo

Answer 1

[José-Luis]

one = rand(100,1);
two = rand(100,1);

nOne = 10;
nTwo = 15;

figure
histogram(one,nOne);
hold on;
histogram(two,nTwo);

%clearer

figure
ksdensity(one);
hold on;
ksdensity(two)

[Flo]

That actually works, but instead of having those kind of graph I'd like to control the bin sizes independently. would you know on how to do that ?

I keep your solution for now :)

thx

[José-Luis]

Please read the documetation for histogram. Sounds like you might want to define custom edges.

[Flo]

I'll have a look on that (already did to be honest). My main concern was to know whether it was feasible or not to keep the 'grouped' option with my kind of problem.

[José-Luis]

I'm not sure I follow. How do you want your plot to look?

[Flo]

I'd like my plot to follow the presentation of the one I posted in the first place (grouped) if such a thing is possible.

[José-Luis]

How would the grouping work if your custom edges have different sizes?

[Flo]

That was my very first question :)

[José-Luis]

Such a plot might be a mess depending on how you bins intersect though...

[Flo]

By looking on google I've found a way apparently. It may interest some person. Let me know what you think about it:

figure(1);
[dummy, t] = hist([one;two], numBin);
nx = hist(one, t); % Sort x into bins.
nx = transpose(nx/sum(nx));
ny = hist(two, t); % Sort y into bins.
ny = transpose(ny/sum(ny));
bar(t, [nx, ny]);

Answer 2

[Duncan Po]

histogram/histcounts have builtin support for Freedman-Diaconis rule:

histogram(x, 'BinMethod', 'fd')

or

[n, binedges] = histcounts(x, 'BinMethod', 'fd')