Sum five elements before a specific number

Hi, I used "findpeaks" to find the IDs of all the peaks in my data. I need for each peak to find the average of five elements before it. the problem is that some of the peaks don't have five elements before it (say the second value is a peak, therefore, there is only one value before it). How can I tell matlab to average whatever is available but not more than 5 elements?

[James Tursa]

By checking the indexes before using them. Please post a code snippet of how you are currently doing this, and then maybe we can offer advice on how best to do the checking.

[Karoline Qasem]

NP=6;
[pks,ID]= findpeaks(X,'NPeaks',NP); 

ID =

      74
      18
       6
       2
      31
       9

These are my peaks

[Image Analyst]

Please attach your data. I'm puzzled as to why your indexes are not in ascending order. I need to see the data to learn why. When I run it, ID always has increasing index values. In the meantime, check out movmean() to get a filtered vector, then pick out the values of the filtered array at those indexes.

[Karoline Qasem]

The data is very huge and messy. The reason why it is going decending is because I used this:

NP=6;
[pks,ID]= findpeaks(Q,'NPeaks',NP,'SortStr','descend');

Answer 1

[Walter Roberson]

data_before = Yourdata(max(1, idx-5) : idx-1)

Note that in the case that idx is 1, this will give the empty result, as "empty" is all of the data that is available before the location.

[Walter Roberson]

Here is a way to do it without loops.

T = [nan(5,1); X(:)];
lenT = length(T);

idx = sub2ind([lenT, 1], bsxfun(@plus, ID, -5:-1), ones(lenT, 5));
result = mean( T(idx), 2, 'omitnan' );

That last line requires R2017a or later. If you have an earlier version and have the stats toolbox you can instead use

result = nanmean( T(idx), 2 );

[Karoline Qasem]

Thanks for the help, it gave an error saying "Error using sub2ind (line 43) Out of range subscript."

[Walter Roberson]

T = [nan(5,1); X(:)];
lenT = length(T);
idx = sub2ind([lenT, 1], 5+bsxfun(@plus, ID, -5:-1), ones(length(ID), 5));
result = nanmean( T(idx), 2 );

[Karoline Qasem]

Amazing! Thanks a lot!