How can I get an average of different matrices stored as separate .mat files?

I have 24 different matrices of dimension 68X68, stored in a folder. I need to create a new matrix that is an average of all those matrices. Using previous FAQ from this forum, I tried :

Num = 24

for i = 1:Num

files = dir(sprintf('ii_Sub*.mat',i))

for k = 1:length(files)

fprintf('Current file : %s\n', files(k).name)

a = load(files(k).name)

meanmatrix = mean(a,24)

end

end

This is creating a 1X1 struct only, while the mean command shows an error as well. How best can I go about it?

[Stephen23]

Your code has sevaral feature that do not make any sense. For example:

• you use nested loops, but never use the outer loop variable i anywhere useful, so it just repeats exactly the same thing 24 times.
• the loop variable i remains totally unused here: sprintf('ii_Sub*.mat',i). There is nothing in that formatting string that uses any input variable.
• The call mean(a,24) makes no sense: a is a cell array, so taking mean of it is not defined. Even if it did work your array have 24 dimensions, so telling MATLAB to sum over the 24th dimension (which will be singular) serves no purpose whatsoever.

[Stephen23]

@Native: please do not close threads that have already have answers.

[Native]

Sure, I was not aware of that. Closed it thinking that the purpose of the question was solved..

As for your earlier comment, thanks for pointing out the fallacies. I must admit that i have started to know the language of code only for a month now, so work in progress. Hope to learn a lot from this forum! :)

Answer 1

[Stephen23]

Assuming that each file contains exactly one variable (a matrix) and that all matrices are the same size and type:

S = dir('*.mat');
N = numel(S);
C = cell(1,N);
for k = 1:N
    T = load(S(k).name)
    C(k) = struct2cell(T);
end
M = cat(3,C{:})
mean(M,3)

[Native]

Hey, this has worked without an error. I'd like to know the logic behind what you did. How can I check that the ans obtained is indeed the average of all the matrices?

Thanks a lot :)

[Stephen23]

"I'd like to know the logic behind what you did."

Sure, here is each line explained:

S = dir('*.mat');            % find all |.mat| files in the current directory.
N = numel(S);                % count those files.
C = cell(1,N);               % preallocate a cell array for the loaded data.
for k = 1:N                  % loop over the number of files. 
    T = load(S(k).name)      % load file data.
    C(k) = struct2cell(T);   % convert structure to cell, allocate to cell array.
end                          % end of loop
M = cat(3,C{:})              % concatenate all MxN matrices into MxNxP array
mean(M,3)                    % mean along third dimension P, giving MxN matrix

"How can I check that the ans obtained is indeed the average of all the matrices?"

Exactly the same way that you would check any other tool or process: by hand yourself, or by using another method. You can easily go through the files, pick the same element from each file (e.g. the first one) and calculate the mean yourself.

"Thanks a lot :)"

Remember to accept my answer if it helped you! Accepting an answer is the easiest way to thank the volunteers who help you on this forum.

[Native]

Thanks, for making the time to do this. That was helpful indeed :)

[Hanisah noh]

hello, this answer helps me.

i have 143 .mat file with 140x122x40 dimension, and i need the average value for all the .mat files

however how do i obtain the MxNxp mean of all the .mat files ?

should i remove/edit the concatenate part (?)

M = cat(3,C{:})              % concatenate all MxN matrices into MxNxP array
mean(M,3)                    % mean along third dimension P, giving MxN matrix

thank you !

[Stephen23]

@Hanisah noh: because your input arrays are aready 3D you can simply concatenate and take the mean along the 4th dimension:

A = cat(4,C{:});
M = mean(A,4)

[Hanisah noh]

@Stephen Cobeldick thank you it works !

[João Vítor Batista Pires Santos]

@Stephen CobeldickThis answer helps me too. But i have 2039 .nc files (17x17x4) and i need the average value for all of them.

The function "struct2cell" doesn't work because the inputs are of double type. How could I turn the structure into cells to take average?

S = dir('*.pgrbl.grb2.nc'); % 2039x1 struct

N = numel(S); % 2039

C = cell(1,N); % 1x2039 cell

for k = 1:N

T = ncread(S(k).name,'PRMSL_L101'); %PRMSL_L101 = Pressure (17x17x4)

C(k) = struct2cell(T);

end

[Stephen23]

@João Vítor Batista Pires Santos: you could try something like this:

for k = 1:N 
    C{k} = ncread(S(k).name,'PRMSL_L101'); % PRMSL_L101 = Pressure (17x17x4)
end
A = cat(4,C{:});
M = mean(A,4)

https://www.mathworks.com/help/matlab/matlab_prog/access-data-in-a-cell-array.html

[João Vítor Batista Pires Santos]

@Stephen Cobeldick It works perfectly!! Thank you!