histogram of a m-by-n matrix

I have a m-by-n matrix of values between 10^-20 and 10^-8 I drew it by

hist(My_mat(:),15);

I obtain only one big bin of the hist, instead I would like to get several of bins between min and max values. Is it correct my code above ? Do you have any suggestions?

Answer 1

[Star Strider]

I would do one of two approaches, depending on what your data distribution is:

1. Increase the number of bins, perhaps from 25 to 100 depending on what you want to do;
2. Take log10(X) and use that instead.

The first is probably the easiest. Experiment until you get the result you want.

[Stefano]

I think that the second one solution may be my way:

1. increasing the number of bins the situation is stable: only decreases the width of the bin

2. several bins are created but in x axis there are negative values (because of lo10 I think), instead all the values are positive, maybe I should modify the x axis values? How to do?

[Star Strider]

I created the data to be plotted as log10Data so assume it is log10 of representative data similar to yours. I am keeping it here for illustration, but you would create it as:

log10Data = log10(My_mat(:));

then use it without other changes being necessary:

log10Data = -randi(12,1,25)-8;

figure(1)
hist(log10Data,15);                             % Plot histogram
hxt = get(gca, 'XTick')                         % Get 'XTick' values
set(gca,'XTickLabel','');                       % Clear default labels
ypos = min(ylim) - diff(ylim)*0.05;             % Set y-position for ‘text’
for k1 = 1:length(hxt)
    xtklbl{k1} = sprintf('10^{%d}',hxt(k1));    % Create x-tick labels
end
text(hxt, ones(1,length(hxt))*ypos, xtklbl, 'HorizontalAlignment', 'center')

[Stefano]

wonderful! I was looking for this solution! How could I to set only some Xticklabel? I thought to display only min,mean and max of hxt, or any other solution to reduce the number of the Label.

[Star Strider]

My pleasure!

That actually was easier than I thought it would be.

The revised code:

log10Data = -randi(12,1,25)-8;

mnD = mean(10.^log10Data);                      % Mean of original data
L10mnD = log10(mnD);                            % Log10 of mean

figure(1)
hist(log10Data,15);                             % Plot histogram
hxt = get(gca, 'XTick')                         % Get 'XTick' values
set(gca,'XTickLabel','');                       % Clear default labels
ypos = min(ylim) - diff(ylim)*0.05;             % Set y-position for ‘text’
hxt = [min(log10Data)  L10mnD  max(log10Data)]; % Set ‘hxt’ to be [min mean max]
for k1 = 1:length(hxt)
    xtklbl{k1} = sprintf('10^{%2.0f}',hxt(k1));   % Create x-tick labels
end
text(hxt, ones(1,length(hxt))*ypos, xtklbl, 'HorizontalAlignment', 'center')

[Stefano]

Ok, thank you so much! If I would to do the hist only of the nonzero values? The zero values maybe insignificant for my interest. I refer always to a matrix

[Star Strider]

My pleasure!

This should do what you want:

log10Data = -randi(12,1,25)-8;

[bcnts,ctrs] = hist(log10Data,15);              % Create histogram counts
bcnts = bcnts(bcnts>0);                         % Eliminate counts with bcnts = 0
ctrs = ctrs(bcnts>0);                           % Eliminate bins with bcnts = 0

ydat = 10.^ctrs.*bcnts;                         % Create new data for calculations
L10YData = log10(ydat);                         % Log10 of new data
mnD = mean(ydat);                               % Mean of edited data
L10mnD = log10(mnD);                            % Log10 of mean

figure(1)
bar(ctrs, bcnts);                               % Plot histogram
set(gca,'XTickLabel','');                       % Clear default labels
ypos = min(ylim) - diff(ylim)*0.05;             % Set y-position for ‘text’
hxt = [min(L10YData)  L10mnD  max(L10YData)]; % Set ‘hxt’ to be [min mean max]
for k1 = 1:length(hxt)
    xtklbl{k1} = sprintf('10^{%2.0f}\n%s',hxt(k1));   % Create x-tick labels
end
text(hxt, ones(1,length(hxt))*ypos, xtklbl, 'HorizontalAlignment', 'center')

It may look a bit strange because of the way I had to recalculate your data with the zero-count bins eliminated. That is done in the ‘ydat’, and creates a data vector with the values of the data at the bin centers weighted by the counts. This skews the data, but is the only way I could calculate it and make the data work for the plot, because with the zero-bins included, the xtick labels did not work correctly. The outcome is that the data plotted is not the data you started with.

Answer 2

[Image Analyst]

You should have 15 bins, though some of them might be virtually empty and too short to notice on the display. So do this:

counts = hist(My_mat(:), 15);
bar(counts, 'BarWidth', 1);
message = sprintf('You have %d bins', length(counts));
uiwait(helpdlg(message));

What does this say? So, if you do in fact actually have 15 bins then you might have to scale the data if you want to see all of the bins, like the others suggested. You can either scale the data values, or the counts - whatever it takes to see your data better.