Only returning NaNs when trying to do a double for loop

I have a matrix of size (361,361) named sivTotEASE.

At specific index i.e. sivTotEASE(m,n) I want to compute the mean (omiting NaNs) of the 8 closest neighbors of sivTotEASE(m,n) and replace the computed value at the location of sivTotEASE(m,n).

I try to do this in a double for loop (first loop for the index corresponding to the rows and second loop for the index corresponding to the columns) but I end up with a matrix containing only NaNs. I don't even have the original values of my matrix anymore... It is just NaNs everywhere...

When I try to compute it manually I do get a real value. For exampe, for index m = 1000, n = 1000 I get the value 8.0114 and not NaN.

Here is my code:

%Load data (see attached files for data)
load('SICnoSIVrow.mat'); 
load('SICnoSIVcol.mat');
load('sivTotEASE.mat');

%Mean of the 8 closest neighbors for sivTotEASE(m,n)
sivTotNeighbor              = sivTotEASE;
for     m                   = SICnoSIVrow
  for   n                   = SICnoSIVcol
        NeighborIt          = nanmean(sivTotEASE(m-1:m+1,n-1:n+1),'all');
        sivTotNeighbor(m,n) = NeighborIt;
  end
end

%Manual test (the following line gives 8.0114)
%test = nanmean(sivTotEASE(m(1000)-1:m(1000)+1,n(1000)-1:n(1000)+1),'all')

Can anyone help me figure out what is the issue?

Thank you

****ACCEPTED ANSWER (see comments)****

%Edited code:

%Load data (see attached files for data)
load('SICnoSIVrow.mat'); 
load('SICnoSIVcol.mat');
load('sivTotEASE.mat');

%Compute neighbor's average
sivTotNeighbor              = sivTotEASE;
for     ii                  = 1:length(SICnoSIVrow)
  for   jj                  = 1:length(SICnoSIVcol)
        m=SICnoSIVrow(ii);
        n=SICnoSIVcol(jj);
        NeighborIt          = mean(sivTotEASE(m-1:m+1,n-1:n+1),'all','omitnan');
        sivTotNeighbor(m,n) = NeighborIt;
  end
end

Answer 1

[Amit Bhowmick]

use this

mean(sivTotEASE(m-1:m+1,n-1:n+1),"omitNan")

[A LL]

Thank you Amit Bhowmick for your quick answer.

I tried the following:

sivTotNeighbor              = sivTotEASE;
for     m                   = SICnoSIVrow
  for   n                   = SICnoSIVcol
        NeighborIt          = mean(sivTotEASE(m-1:m+1,n-1:n+1),'all','omitnan');
        sivTotNeighbor(m,n) = NeighborIt;
  end
end

But it still only returns NaN. Do you know how to fix this?

Thank you

[Amit Bhowmick]

As per the attached file this variable is missing ('SICnoSIVcol'). So I can not check or verify.

[A LL]

This is my mistake, I edited my question with the proper variables.

[Amit Bhowmick]

All the values of variable load('sivTotEASE.mat') contains NaN.

[A LL]

There are indeed NaNs in the first rows and columns but it is not all NaNs

notnan = find(~isnan(sivTotEASE));
%notnansize = size(notnan) %returns 18115

There are in fact 18115 entries with non nan values

And if you use the following command without the NaN flag (as in your picture) you will indeed get a NaN instead of the value 8.0114 because the value is surrounded by some NaNs.

test = mean(sivTotEASE(m(1000)-1:m(1000)+1,n(1000)-1:n(1000)+1),'all','omitnan')

[Amit Bhowmick]

within the loop what is m and n ?

Are they scaler or vector

[Amit Bhowmick]

I think the following correction required

for     ii                  = 1:length(SICnoSIVrow)
  for   jj                  = 1:length(SICnoSIVrow)
        m=SICnoSIVrow(ii);
        n=SICnoSIVrow(jj);

[A LL]

It is the index of rows (m) and columns (n) where I want to compute de average of the neighbors.

m is the vector from the variable SICnoSIVrow.mat and n is the vector from the variable SICnoSIVcol.mat

m                   = SICnoSIVrow;
n                   = SICnoSIVcol;

[A LL]

I think I solved my issue by writing the following:

%Load data (see attached files for data)
load('SICnoSIVrow.mat'); 
load('SICnoSIVcol.mat');
load('sivTotEASE.mat');

%Compute neighbor's average
sivTotNeighbor              = sivTotEASE;
for     m                   = [1:length(SICnoSIVrow)] %row
  for   n                   = [1:length(SICnoSIVcol)] %col
        NeighborIt          = mean(sivTotEASE(SICnoSIVrow(m)-1:SICnoSIVrow(m)+1,SICnoSIVcol(n)-1:SICnoSIVcol(n)+1),'all','omitnan');
        %disp(NeighborIt);
        sivTotNeighbor(SICnoSIVrow(m),SICnoSIVcol(n)) = NeighborIt;
  end
end

%test = mean(sivTotEASE(SICnoSIVrow(1000)-1:SICnoSIVrow(1000)+1,SICnoSIVcol(1000)-1:SICnoSIVcol(1000)+1),'all','omitnan');

[A LL]

I missed your comment but I indeed think you are right with your suggestion to edit my code but jj and n would have to be for the variable SICnoSIVcol (and not for SICnoSIVrow):

for     ii                  = 1:length(SICnoSIVrow)
  for   jj                  = 1:length(SICnoSIVcol)
        m=SICnoSIVrow(ii);
        n=SICnoSIVcol(jj);

This is also what I ended with (see comment above)!

Thank you

[Amit Bhowmick]

Welcome