calculate the mean of several columns

Hi all, I am new to Matlab and I am struggling with this problem. I have a matrix with 112 columns and 100 rows. How can I calculate the mean of column 1-7 , 8-14, 15-21,...., 106-112 and put the a new matrix with 16 columns and 100 rows. I appreciate your help Sobhan

[Andrei Bobrov]

eg

A = rand(100,112);

fun = @(block_struct)mean(block_struct.data,2);

out = blockproc(A,[100,7],fun);

[Walter Roberson]

KAUST comments "The best answer"

Answer 1

[Wayne King]

One way is simply with a for loop

A = randn(100,112);
for nn = 0:15    
    B(:,nn+1) = mean(A(:,nn*7+1:(nn+1)*7),2); 
end

[Oleg Komarov]

I bet this is more efficient than the one liner and definitely more readable.

Answer 2

[Oleg Komarov]

A = rand(100,112);
B = reshape(mean(reshape(A.',7,[])),16,100).';

% Check
isequal(mean(A(2,8:14)),B(2,2))

Answer 3

[Thomas]

You could reshape your matrix with 7 columns becoming 1 in the new matrix and taking the mean of each column

Eg.:

c=rand(4)     % 4x4 matrix
d=reshape(c,[],2)   % take two colm and form 1 total 2
p=mean(d)        % mean of each columns

In your case

    c=rand(100,112)     % 4x4 matrix
    d=mean(reshape(c,7,[]))  % form 1600 columsns
    final_out=(reshape(d,16,100))    % mean of each columns

[Oleg Komarov]

This solution will give 1 x 16, whereas the OP wants 100 x 16. The average is across blocks AND rows.

[Thomas]

thanks, edited accordingly..

[Oleg Komarov]

Now, your mean is across the first 7 rows (first column). You have to transpose and re-transpose back. Honestly, in this cases I think Waynes solution is best in terms or readability and performance.

Answer 4

[Sobhan]

Dear all, Waynes` solution worked like a charm. You are awesome. Thanks a lot for your nice suggestions :)