Calculate running average with NaN

Hello,

I have an array similar to array A below, but much larger:

A = [1:15;12 15 20 21 12 14 12 23 41 21 12 11 33 21 42;5 NaN NaN 42 32 42 33 22 12 NaN NaN NaN 12 42 11]';

I would like to calculate the running average of the 3rd column with a window of 4 data points and plot it overlaying on top of the original data. However, I couldnt get it to calculate the correct running average.

RA = [];
for j = 1:(length(A)-4)
    RA(j,1) = mean(A(j:j+4,1));
end

figure
plot(A(:,1),A(:,2));
hold on
plot(A(:,1),A(:,3));
plot(A(:,1),RA);

How can I find the RA and overlay it on top of the old data correctly?

Thanks, Max

Answer 1

[Guillaume]

If using 2016a or later, simply:

RA = movmean(A, 5, 'omitnan', 'EndPoints', 'discard');

[Max Bernstein]

I dont have movmean, is there a similar script somewhere?

Answer 2

[Image Analyst]

Try this:

A = [1:15;12 15 20 21 12 14 12 23 41 21 12 11 33 21 42;5 NaN NaN 42 32 42 33 22 12 NaN NaN NaN 12 42 11]'
% Extract column 3 only.
col3 = A(:, 3)
% Find out where the nans are.
nanIndexes = isnan(col3)
% Turn nans into zeros
col3(nanIndexes) = 0
kernel = [1,1,1,1]; % Kernel to look at 4 elements.
% Do a running sum.
sums = conv(col3, kernel, 'valid')
% Do a running count.  Count how many non-nans there are.  In other words, how many good numbers.
counts = conv(double(~nanIndexes), kernel, 'valid')
% Divide them to get the means.
runningMeans = sums ./ counts

runningMeans =
                      23.5
                        37
          38.6666666666667
                     37.25
                     32.25
                     27.25
          22.3333333333333
                        17
                        12
                        12
                        27
          21.6666666666667

[Max Bernstein]

I had to change the 'valid' to 'same' in order to get the same array size to plot. However, this method seems to shift the magnitude of the running average down for some reason.

[Image Analyst]

With a window size of 4, there are 1 or 2 elements outside when the 'same' option is used. Normally the window size is odd anyway. But when the window is at the far left, about half the window will overlap the array and half will be "off" the far left end of the array. Same for the right side. You have to decide how you want to handle what's called "edge effects" which is what happens when the moving window starts to move off the main array.