gelman_rubin returns wrong values about pymc HOT 3 CLOSED

Its not clear to me what the bug is here. The G-R statistics should be approximately one if convergence is reached. They are based on a stochastic sample, so you should not expect them to be identical each time you run the code. If I am missing something, please clarify.

from pymc.

y is a deterministic node and it's equal to x[2], so the trace of y and the trace of x[2] is identical,
which means that GR of x[2] should be equal to the GR of y.
In the result above you can see that x[2] has GR of 0.949880...
while the GR of y is 1.01808

from pymc.

Got it, thanks. It was just a silly transpose error. Fixed in 81999ab

from pymc.