[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
## McSim query

**From**: |
Suhrid Balakrishnan |

**Subject**: |
McSim query |

**Date**: |
Fri, 25 Oct 2002 02:46:42 -0400 (EDT) |

Hi,
I'm trying to do some analysis with McSim 4.1 and I'd like to specify
priors in a heirarchical fashion exactly as the authors of the software
did in their 1996 paper - Bois, F. Y., Gelman, A., Jiang, J., Maszle, D.,
Zeise, L. and Alexeef, G. (1996). Population toxicokinetics of
tetrachloroethylene. Archives of Toxicology 70:347-355. also used in
Gelman, A., Bois, F. Y. and Jiang, J. (1996). Physiological
pharmacokinetic analysis using population modeling and informative prior
distributions. Journal of the American Statistical Association
91:1400-1412.
The problem I'm facing is that I see no way to easily incorporate the
Lognormal hierarchical prior (mu*, Sigma*) (where mu* and Sigma* are the
geometric mean and standard deviations respectively = exp(mu) and
exp(Sigma)) with further conditional distributions on mu ~ Normal (M,S)
and Sigma as inverse-gamma(a,b) (a, b being the shape and scale
parameters, respectively) in the McSim framework.
I've tried a number of things including toying with levels, exponents,
outputs, scaling etc. and nothing seems to work.
What one desires, I presume, is a way to specify the following
hierarchical distributions in the simulation file (which the authors used
in both references above):
Distrib (Sigma_V, InvGamma, 2, 0.1823);
Distrib (Mean_V, Normal, 0.1, 1.2); # values are abitrary
Distrib (V, TruncLogNormal, exp(Mean_V), exp(Sigma_V), 1e-6,10);
But the above formulation isn't allowed by McSim (no exponentiation
alllowed in the Distrib specification).
Or an alternative way to get the same effect - is there something obvious
that I'm not seeing? Presumably, there is a way to do this, I would be
greatly appreciative if anyone would let me know how.
Thanks in advance,
Sincerely,
Suhrid

[Prev in Thread] |
**Current Thread** |
[Next in Thread] |

**McSim query**,
*Suhrid Balakrishnan* **<=**