Average positive numbers in a row

Question

I have a numpy matrix, and each row has a combination of positive and negative numbers.

I want to create a new vector, that gives me the average of all the positive numb

Answer 1

A possible solution would be

• select all non-negative
• compute explicitly the mean as sum/count

In code:

notneg = x >= 0
result = (x * notneg).sum(1) / notneg.sum(1)

Answer 2

If it's guaranteed to have at least one positive number (>=0) per row, you could convert the negative numbers (excluding 0) to NaNs with np.where and then use np.nanmean along the rows, like so -

np.nanmean(np.where(A>=0,A,np.nan),axis=1)

Sample run -

In [69]: A
Out[69]: 
array([[ 2,  3, -6, -6, -4],
       [-5, -6, -1, -1,  3],
       [-8,  5, -7, -9, -9],
       [-3,  0,  7, -5, -6]])

In [70]: np.nanmean(np.where(A>=0,A,np.nan),axis=1)
Out[70]: array([ 2.5,  3. ,  5. ,  3.5])