Yeah, giving some sane starting values seems to have worked.
{'95% HPD interval': array([[ 3.61151514, 5.14205003],
[-4.13174923, -2.64597387],
[ 2.50344411, 3.85207883],
[ 0.85618149, 2.28345262]]),
'mc error': array([ 0.03873973, 0.0391429 , 0.03033799,
0.03350415]),
'mean': array([ 4.38028987, -3.36625853, 3.28085591, 1.61798147]),
'n': 5000,
'quantiles': {2.5: array([ 3.6564718 , -4.13003482, 2.52034961,
0.85618149]),
25: array([ 4.11925573, -3.63458265, 3.05276479, 1.389134 ]),
50: array([ 4.32780316, -3.33692739, 3.30093981, 1.62899525]),
75: array([ 4.62855652, -3.04346921, 3.50765626, 1.87653607]),
97.5: array([ 5.29443095, -2.64191083, 3.90125011, 2.28345262])},
'standard deviation': array([ 0.41222244, 0.41570342, 0.34188305,
0.36581739])}
I made a few other cosmetic changes as well. Here is the updated
model:
http://dl.dropbox.com/u/233041/mvreg.py
Cheers,
Chris