Accumarray error for logspace binning

Hi all,

I'm trying to bin some of my data according to a nonuniform spacing method based on a logarithmic scale, whereby I create 100 bins from the minimum value to the maximum value. First I use histc function to generate bin indices, then I try using accumarray to build the matrix based on the mean value as follows:

[unused, binidx] = histc(M4M2_2, ...
  logspace(log10(min( M4M2_2 )), log10(max( M4M2_2) ),100+1));
binidx( isnan( M4M2_2 ) ) = [];
binidx = uint8( binidx );
M4M2_2n = M4M2_2; M4M2_2n( isnan( M4M2_2n ) ) = [];
means = accumarray(binidx, M4M2_2n, [], @mean);

However, this does not work even after trying to make the binidx variable into an integer variable using uint8. I still get the following error, which I cannot figure out because I checked that all the values in the binidx matrix were in fact positive integers:

Error using accumarray
First input SUBS must contain positive integer subscripts.

[Matt J]

I checked that all the values in the binidx matrix were in fact positive integers:

Attach a .mat file containing binidx and M4M2_2n so we can convince ourselves of that.

[Drew Mahedy]

Try now

Answer 1

[Star Strider]

I can’t follow what you are doing. If you want to plot binned data logarithmically, this code from an Answer I provided a couple years ago may do what you want:

x = 10+2*randn(1, 100);                                             % Create data

bins = linspace(floor(min(x)), ceil(max(x)), 25);   % Define bins
xc = histc(x,bins);                                                 % Do histogram counts

figure(1)
bar(log(bins), xc)                                                  % Plot bars against log of ‘bins’
logxts = get(gca, 'XTick');                                         % 'XTick' values currently logarithmic
expxts = exp(logxts);                                               % Take antilog to use them as new ‘XTickLabels’
set(gca, 'XTickLabel', floor(100*expxts)/100)

You will probably have to experiment with it to get the result you want.

Answer 2

[Roger Stafford]

If the line

   M4M2_2n = M4M2_2; M4M2_2n( isnan( M4M2_2n ) ) = [];

succeeds in shortening M4M2_2, then the required “one-to-one” matching between binidx and M4M2_2n is violated. Conceivably this could cause a misleading error message such as you received. Here is a quote from the ‘accumarray’ site:

http://www.mathworks.com/help/matlab/ref/accumarray.html

“In both cases, a one-to-one pairing is present between the subscripts in each row of subs and the data values in val.” In any case, it is a mismatch that you surely would want to correct.

[Drew Mahedy]

The fact the binidx line comes first means that binidx gets shortened based on M4M2_2 before M4M2_2 gets shortened based on itself.

Answer 3

[Matt J]

Your binidx values are non-negative, but are not positive as you claim:

>> min(binidx)

ans =

      0