ignore NaN values in loop

Hi everyone,

So I have a numeric matrix "walknov" with several columns of different lengths and I interpolated them all to be 100 points and created a new matrix "CC" with the following loop:

for i = 2:16
   A = walknov(:,i);
    A(any(isnan(A),2),:) = []; 
    A = interpft(A, 100);
    CC = [CC A];
end

NaNs fill the differences of length between columns, so I have eliminated each row containing NaN everytime I interpolate each column. The problem is that some columns have no values at all (i.e. only NaNs) and this makes the loop stop as the interpolation function won't accept such values.

I do not want to drop the column as I need the matrix to match other matrices, but instead I would like to just leave NaN values in those missing columns. Is there a way to solve this?

Thanks a lot!

PS: I have attached a sample dataset

Answer 1

[Walter Roberson]

if isempty(A)
    CC(:,i) = nan;
else
    A = interpft(A, 100);
    CC(:,i) = A;
end

Can interpft operate with only one non-nan value?

[Cristina Elmeua]

This has worked great Walter. A very simple and effective approach, thanks.

Answer 2

[David Hill]

CC=[];
for i = 2:16
   A = walknov(:,i);
   A = interpft(A(~isnan(A)), 100);
   if isempty(A)
       CC=[CC;nan(1,100)];
   else
       CC = [CC;A];
   end
end

[Cristina Elmeua]

Thanks for the help David!

Answer 3

[Adam Danz]

The image below shows the location of your missing values. The good news is that the missing values either consume entire columns or the end of columns instead of being dispersed.

To work around this,

for i = 2:16
    A = walknov(:,i);
    A(any(isnan(A),2),:) = []; 
    if isempty(A)
        A = NaN(size(CC,1));  % 1 column?
    else
        A = interpft(A, 100);
    end
    CC = [CC A];
end

This assumes CC is defined before the loop. Otherwise if the empty 'A' appears on the first loop, there will be an error indicating that CC is not defined.

Show NaN pattern.

imagesc(isnan(w2_2))  % or heatmap(double(isnan(w2_2))

[Cristina Elmeua]

Thanks for the help Adam!