Add large amount of variables

Hello, I have a set of variables which are all Matrixes and their name is usually "Hsig_day_time". I would like to add them all but it there are about 120 and possibly there will be more than 1000.

Is there a way to add them all without writting each variable name in a line?

[Geoff Hayes]

@Magnus Schneider - where do you need to add the variables? Do you mean to add them to a cell array?

[Magnus Schneider]

Basically I want to add every Matrix , and then divide it by the number of Matrixes that I have. I want to have the average of every element of those Matrixes.

Ex

A= 1 2 B=5 6 C=9 10

3 4 7 8 11 12

Objective= (A+B+C)./3

[Geoff Hayes]

@Magnus Schneider - are your matrices contained within a cell array or other matrix (assuming all are of the same dimension) or do you have individually named variables for each matrix? If the latter, then I recommend that you transfer them all to a single muti-dimensional array to facilitate the matrix addition and avoid issues that occur when you have individual variables for each of your matrices (like the 150 in this case).

[Magnus Schneider]

@Geoff Hayes They are all inside a .mat file. However, I think they are individually named. Is it possible to create the multi-dimensional array without having to write each variable name?

[Rik]

You can load to a struct and then assign the loaded data to whatever array makes most sense. Writing variable names explicitly is generally a bad idea and should be avoided if possible.

[Magnus Schneider]

@Rik These variables are created by an external software. I am not sure if it is possible to load them as a struct. At least I can't do it.

[Stephen23]

"I am not sure if it is possible to load them as a struct. At least I can't do it."

https://www.mathworks.com/help/matlab/ref/load.html

Answer 1

[dpb]

OK, I'll just make another Answer so previous conversation doesn't distract

Brute force; may be something a little more clever possible, but...

S=load('IBERIA.mat','-regexp', '^Hsig');        % load the variables beginning with Hsig only
C=struct2cell(S);                               % turn the fields into cells in cell array
A=C{1};                                         % begin accumulator for sum
for i=2:numel(C)                                % and iterate over the size of C
  A=A+C{i};                                     % sum each array
end
mnA=A/numel(C);                                 % compute mean

Above doesn't deal with NaN -- I noticed there are at least some matrices with NaN in them; I didn't search for which or where. I did note that the first 10 locations in both directions are finite.

You'll know better what to do with those locations...

[Magnus Schneider]

Thank you so much! it worked!

[Image Analyst]

@Magnus Schneider can you please click the "Accept this answer" link to award @dpb "Reputation Points" and to let others know it's been solved. Thanks in advance. 🙂

[Magnus Schneider]

yes of course.

Answer 2

[dpb]

Don't try to explain in words; show us precisely what the external software generates.

" They are all inside a .mat file..."

Then at a minimum show us what

whos -file TheFileName.mat

returns for one of these files.

Even better would be to attach one here..."help us help you"

The initial (following) was written before I saw the above comment so presumed each was a separate file.

If they are 2D arrays and the dimensions are not humongous for each, then you can simply read each into a 3D array where each file is a plane and use mean(A,3)

If the total won't fit in memory, then a mean is simply a sum which can be computed as a running summation with only one file at a time, then divide the sum by the total number of files.

Either is a basically trivial task to iterate over all files in a particular location.

[Magnus Schneider]

I am sorry this is the first time I am posting here. The file is in the attachment. The final objective is to obtain a Matrix with the averages of each element of the Hsig matrixes. In the real examples there will be hundreds of Hsig matrixes...