Skip subs/index values with accumarray

I created an index (groupID) for a vector A. The vector is sorted from the smallest value to the biggest one. The function and the groupID look like:

[~,groupID]=histc(A,0:5:100);
groupID = [1 1 3 3 3 5 5];

In the next step I want to calculate the average for every groupID. I did this as follows:

groupMeans = accumarray(groupID,A,[],@mean);

My problem is that accumarray fills in the output array for every skipped groupID (because e.g. 2 and 4 are missing in the index) a zero. I dont want these zeros in my output array. Is there a solution for this? My Matlab version is R2012b

Answer 1

[Stephen23]

Use unique's third output to get the indices:

[uni,~,idx] = unique(groupID(:));
groupMeans = accumarray(idx,A,[],@mean);
[uni,groupMeans]

[leonidas86]

With unique I get hundreds of groupIDs.. This isnt nice for plotting with the bar function. Can I change the number of groupIDs?

[Walter Roberson]

Increase the width of your bins to reduce the number of group IDs.

[leonidas86]

How can I do this?

[Stephen23]

"How can I do this?"

The histc help explains how the second input defines the bin ranges. If you want to change the bin ranges to make wider bins then you need to change the vector that you used (to have a larger step size).

[Walter Roberson]

In the original posting you used

histc(A,0:5:100);

This would create 21 bins, [0 5 10 15 20 ... 100]. To increase the width of your bins in this case you would use an increment larger than 5, such as 0:8:100 .

Note: more robust would be to use linspace(0, 100, N) where N is the number of bins you wanted.

hist() and histcounts() permit you to pass the number of bins directly.