Mean over specific columns

Hi :)

I have a 450x42 table.

I would like to calculate the mean over cloumns 2:5 7:10 12:15 and so

mean([2:5],2) mean([7:10],2).....

anybody an idea :)

[Walter Roberson]

Provided that the pattern keeps up:

squeeze( mean( reshape(Tab{:,setdiff(1:size(Tab,2),1:5:size(Tab,2))}, size(Tab,1), 4, []), 2))

[Antonio Melieni]

no idea how to usw that.... can u maybe explain it :)

[Walter Roberson]

In the above, Tab would be your table object. All of the variables (columns) in the table would have to be numeric for it to work properly. The code I posted could have problems your variables are not all the same numeric type, such as if columns 2:5 were double but columns 7:10 were uint8.

Answer 1

[John D'Errico]

Easy peasy? Convert to an array first of course. Or, if some of those columns are non-numeric, then convert only the indicated columns to an array.

But it is pretty easy to generate that list of columns.

T = rand(450,42);
M = mean(T(:,[2:5]' + (0:5:35)),1);

So the list of column indices is:

ind = reshape([2:5]' + (0:5:35),1,[]);