[SciPy-user] Bias in numpy.random.multivariate_normal?

[John Reid]

Matthieu Brucher wrote:
> IIRC, you computed the biaised variance, so it is normal to see biais. 
> Try dividing by (sample_size-1), it should be the unbiaised estimator.
> 
That wouldn't change the sign of the answer. In any case I get the same 
results when dividing by sample_size-1.