Find the standard error of the mean for each column in a matrix containing NaN and numerical values

I have a 5x4 matrix with NaN and numerical elements as shown below:

A = [ NaN NaN NaN 3.8905

NaN NaN NaN 3.4095

NaN NaN 3.0360 NaN

NaN NaN 2.5295 NaN ];

I was able to find the mean using nanmean and standard deviation using nanstd for each row but how do I calculate the standard error of the mean for each column in this matrix?

When I use length(A(~isnan(A))), size(A(~isnan(A)),1) and numel(A(~isnan(A))) for the non NaN values in this matrix, I always get the total number of non NaN values as opposed to non NaN values in each column.

Answer 1

[Chris]

1. That's a 4x4 matrix, not 5x4.
2. The default behavior or nanmean and nanstd is to give you the statistic for each column. Taking the standard deviation of each row in this example would give you zeros.
3. To get the number of non-nans in each column:

sum(~isnan(A))

standard error:

nanstd(A)./sqrt(sum(~isnan(A)))

[Leeba Chacko]

Thanks Chris, how do I calculate the standard error of the mean for each column?

[Chris]

I edited my answer.

[Leeba Chacko]

Thanks Chris, this is really helpful!