Loop multiple files from a cell

Hi, I am trying to get the nanmean from my data. Firstly I load my files (.mat) into a cell (files_mat). Now files_mat is (19x1) and each cell has a matrix of (10200x13). I would like to get the nanmean from all matrices and each matrix has different names, so I am doing a Loop and the resultas is going to be in a new matrix (res_est).

res_est = NaN(19,13)
for i=1:length(files_mat)
    res_est(k,13)=nanmean(files_mat{k,1}.Matriz_A(:,1:13)
end

I get the result for the first matrix, but I would ike to get the nanmean of all matrices successively, the problem is: that I have to change manually the name.

for i=1:length(files_mat)
    res_est(k,13)=nanmean(files_mat{k,1}.Matriz_B(:,1:13)
end

Some advice to get the nanmean from all files? Thanks

Answer 1

[Walter Roberson]

Currently your loop to load looks somethng like,

dinfo = dir('*.mat');
nfiles = length(dinfo);
files_mat = cell(nfiles, 1);
for K = 1 : nfiles
    filename = dinfo(K).name;
    files_mat{K} = load(filename);
end

You can instead use

dinfo = dir('*.mat');
nfiles = length(dinfo);
files_mat = cell(nfiles, 1);
for K = 1 : nfiles
    filename = dinfo(K).name;
    data_struct = load(filename);
    fn = fieldnames(data_struct);
    files_mat{K} = data_struct.(fn{1});
end

Now files_mat will be a cell of arrays instead of cell of struct.

If all of the arrays are the same size you could use a numeric array instead of a cell:

dinfo = dir('*.mat');
nfiles = length(dinfo);
files_mat = zeros(0,0,nfiles);
for K = 1 : nfiles
    filename = dinfo(K).name;
    data_struct = load(filename);
    fn = fieldnames(data_struct);
    this_data = data_struct.(fn{1});
    files_mat(1:size(this_data,1), 1:size(this_data,2), K) = this_data;
end

and after that you can nan_mean across entire sections:

res_est(:,13) = nanmean(files_mat, [1 2]);

Notice here that I take the mean over the first two dimensions. Your code

for i=1:length(files_mat)
    res_est(k,13)=nanmean(files_mat{k,1}.Matriz_B(:,1:13)
end

has a confusion between i and k, and is taking the nan_mean() over 10200 rows by 13 columns but assigning to a scalar location, so we can guess that you want to take the mean of everything rather than a row or column mean.

Also since your array has 13 columns and you are accessing (:,1:13) then the (:,1:13) is redundant, which is why I left off the indexing from files_mat.

[Fabian Moreno]

Hello Walter, thaks a lot for help me. I became more familiar with this option. I am new in matlab.

dinfo = dir('*.mat');
nfiles = length(dinfo);
files_mat = cell(nfiles, 1);
for K = 1 : nfiles
    filename = dinfo(K).name;
    data_struct = load(filename);
    fn = fieldnames(data_struct);
    files_mat{K} = data_struct.(fn{1});
end

All I wanted to do was this

stad_mean = NaN(19,6);
for ii=1:19
    stad_mean(ii,:) = nanmean(files_mat{ii,1}(:,8:13)); 
end

You know a lot, I would like to learn more, Do you know where I can see more things. You are my hero, I am so grateful with you. I though the problem was that I was getting the data to a struct and not to a double.

[Walter Roberson]

If you use the version I showed with loading into a numeric array, then you would be able to use

stad_mean = squeeze(nanmean(files_mat, 1));

I though the problem was that I was getting the data to a struct and not to a double

Yes, that was the problem. In particular, you had a struct with a field name that was the same as the name of the variable stored in the file, because when you call load() with an output, it creates a struct with one field for every variable in the file. fieldnames() asks what names were received, and the syntax STRUCTURE.(EXPRESSION) dynamically evaluates EXPRESSION to find the name of the field to access. STRUCTURE.(EXPRESSION) is a special syntax in MATLAB.

[Fabian Moreno]

Thank you again Walter :D, both given solutions worked good.