csr_vcorrcoef calculates wrong correlation

[WeilerP]

Description

In csr_vcorrcoef, the radicals in the denominator of the Pearson's correlation are calculated incorrectly. For 1d arrays X, the radical should be

np.sum((X - X.mean()) ** 2)

or equivalently

np.sum(X ** 2) - 2 * np.sum(X) * X.mean() + len(X) * X.mean() ** 2

The actual implementation is missing the factor 2 as well as len(X).

Versions

scvelo==0.2.4.dev48+gba75c4a.d20210719  scanpy==1.7.2  anndata==0.7.6  loompy==3.0.6  numpy==1.20.3  scipy==1.6.3  matplotlib==3.4.2  sklearn==0.24.2  pandas==1.2.4

[WeilerP]

Closed via #679.