Average IF NaN counting

Hi,

I have a 20*207 matrix which contains numbers and NaN.

I need to average each column if the amount of NaN is <5. If the number of NaN is >=5 then I would like to remove that column from the matrix. How can I do it? I've been trying some stuff but I'm not able to achieve it.

thanks.

[Geoff Hayes]

Afonso - what have you been trying? Please copy and paste your code to this question so that we can get an idea of what you have attempted and determine why it is not working. Presumably you are using isnan to determine which elements in your column are NaN...

Answer 1

[Robert]

Some test data

x = randn(20, 207);
x(randi(numel(x), 800, 1)) = nan;

Identify the NaNs with isnan and count the number per column with sum. Then select the columns that have fewer than 5 NaNs.

good_enough = sum(isnan(x)) < 5;
x = x(:, good_enough);

And use nanmean to take the average of the remaining columns.

avg = nanmean(x);

If you don't have the stats toolbox, try the following

the_nans = isnan(x);
num_nans = sum(isnan(x));
good_enough = num_nans < 5;

x = x(:, good_enough);
the_nans = the_nans(:, good_enough);
num_nans = num_nans(:, good_enough);

x(the_nans) = 0;
avg = sum(x);
avg = avg ./ (20 - num_nans);

[Osnofa]

Thank you Robert! way simplier than I thought.

I tried the 1st method.