How does one plot a histogram from the histogram counts?

I have produced an array of counts, specifically produced with:

histcounts

the reason I do this is because storing the actual data is way to expensive in terms of storage. So on the fly I update my histcount. Now that I have them, how do I plot the histogram?

----

demo code:

W_hist_counts = zeros(1,D);
for i=2:iter+1
    W = get_vector_we_want_stats(W)
    %
    W_hist_counts = W_hist_counts + histcounts(W,nbins);
end

----

http://stackoverflow.com/questions/43058290/what-is-the-official-matlab-way-to-plot-the-values-of-histcounts-into-a-histogra

Answer 1

[the cyclist]

The most straightforward way is probably to use the bar command.

[Brando Miranda]

there is a weird caveat what the number of hist_counts and edges dont agree when returned by hist_counts. Do you know how to fix this?

[Star Strider]

If the edges are regularly spaced, this is what I do to create a vector of centers for the bar plot:

e = 1:5                                         % ‘Edges’ Vector
xt = e(1:end-1)+mean(diff(e))/2                 % ‘Midpoint’ Vector

e =
     1     2     3     4     5

xt =
          1.5          2.5          3.5          4.5

[Brando Miranda]

also, bar doesn't accept normalization options. I know histcounts does but that forces me to have all the data before I feed it to histcounts, which I don't have because I am in the real of big data. I can produce statistics (like counts) on the fly but not store the whole thing in matlab.

[the cyclist]

I assume you want to normalize to an overall sum of 1. You should just be able to plot

W_hist_counts/sum(W_hist_counts)

to normalize, right?

Answer 2

[Image Analyst]

histogram() can plot the histogram data (counts vs values) if you pass in the counts and values instead of the data. Otherwise you can do it manually with bar().