matrix nan

hi, i have multiple nans in my 1000x1 matrix. I want to replace the nans with the average of the number above and below it......please note if there are more than 1 nans in a row it should take average of the numbers above and below it.

FOR EXAMPLE. My Array:

2

3

nan

5

nan

nan

9

3

2

........desired output:

2

3

4

5

7

7

9

3

2

Answer 1

[Nirmal]

a=[2 3 NaN 5 NaN NaN 9 3 2];

[m,n]=size(a);

b=isnan(a);

for j=1:n
    if b(j)==1
        li=j-1;
        while b(li)==1
            li=li-1;
        end
        ui=j+1;
        while b(ui)==1
            ui=ui+1;
        end
        avg=(a(li)+a(ui))/2;
        a(j)=avg;
    end
end

The code however doesnt handle the boundary case. Its just couple of if checks.

Answer 2

[Andrei Bobrov]

data = [2
3
nan
5
nan
nan
9
3
2];

a1 = isnan(data);
vle = mean(data([strfind(a1(:)',[0 1]);strfind(a1(:)',[1 0])+1]));
id = bwlabel(a1);
data(a1) = vle(id(id~=0));

Answer 3

[Image Analyst]

Here's a method that I think is pretty easy to understand and is only 4 lines long.

% Generate sample data.
array2D = randi(5, [20 20])
nanIndices = randperm(numel(array2D))
% Take first 50 of them nan's and assign to the array
array2D(nanIndices(1:50)) = nan
% Now we have our sample data.

% Now we are ready to begin.  Here are the 4 lines of code:
% Find the nan's
nanIndices = isnan(array2D);
% Replace nan's by zeros
array2D(nanIndices) = 0
% Get the average matrix in a vertical 3x1 window.
verticalAverageMatrix = conv2(array2D, [1;1;1], 'same') / 2
% Replace nan's with the average
array2D(nanIndices) = verticalAverageMatrix(nanIndices)