Var(x) and cov(x, x) don't give the same result in numpy
You must use z=cov(x,bias=1) in order to normalize by N ,because var is also norm by N(according to this
The default ddof of cov (None) and var (0) are different. Try specifying the ddof (or bias):
>>> cov(x, x, ddof=0)array([[ 8.25, 8.25], [ 8.25, 8.25]])>>> var(x)8.25