Objet : Help-mcsim Digest, Vol 3, Issue 3

[Frédéric BOIS]

Hi ! I post this to the help mailing list since it may be useful to others

You can do two things: 
A/ set the MCMC statement "printflag" to 1 and give a "restart" file. It will 
read the last line of the restart file 
     (that should be a parameter vector, e.g. the max posterior param. vector) 
and do 1 simulation of all the 
     data point in the MCMC input file with the parameters set at the values 
read in.

B/ more generally you can use the "setpoints" statement to read all parameter 
vectors in a MC or MCMC output file
     and for each one ask for a set of simulations. You can request any output 
you want then. The simulation requested
     will be performed for all the parameter vectors read in.


F. Bois


==========================
Frederic Y. Bois,
Unite de Toxicologie Experimentale, responsable
INERIS
Parc ALATA, PB2
60550 Verneuil en Halatte
FRANCE
tel: + 33 (0)3 44 55 65 96
fax: + 33 (0)3 44 55 66 05
email: address@hidden
web: http://www.ineris.fr 


>>> <address@hidden> Jeudi 13 février 2003 18:02:58 >>>
Today's Topics:

   1. Estimating a posterior predictive distribution (Iago Mosqueira)

Hello,

I would like to know if MCsim allows for the estimation of the posterior
predictive distribution, in order to be able to compare model predicted data
with the observed data.

So, when I define the likelihood function,

Distrib(y, Normal, Prediction(y), .5);

I would be like to be able to say

Distrib(z, Normal, Prediction(y), .5);

where z is an output variable estimated from the predicted value of y at
each PrintStep. This would allow for the estimation of posterior predictive
distributions and Bayesian p-values (Gelman 1995, p.170)

Is there any way of achieving this in MCsim?