Hi to everyone,
I'm a PhD student and I'm involved in non parametric analyses of
hierarchical models. I tried to use package DPpackage on my data, but I
encountered some problems in interpreting ouputs. Can anybody help me?
The problem can be remued as follows: I have a logit hierarchical model
for survival (i.e. binary response) in patients affected by heart
failure (the court consists of n=536 subjects), admitted in J different
hospitals. The idea is to explain survival by means of linear predictor
of relevant covariates with a random effect on grouping factor,
represented by hospital of dmission. The main goal would be the
reconstruction of random effect density, because we are interest in
founding (if present) groups of "similar" hospitals.
Now, I found some trouble in doing this beause I don't understand how
outputs "ss", "ncluster" and "mub" work.
Particullary, how does the
package compute "ss" starting from "ncluster" distribution?
How should I
interpret the vector "mub"? I tried to look into the fortrand code,
but
I'm not used to that language so it didn't help me a lot. Could anybody
make me understand how "ss" and "ncluster" are related, and
how to build
the predictive densities of the random effect starting from "mub"s
elements? It would be a great gain in my work.
Looking forward to hear news from you
thanks a lot in advance
Regards
Francesca Ieva
--
-----
Francesca Ieva
MOX - Modeling and Scientific Computing
Dipartimento di Matematica "F.Brioschi"
Politecnico di Milano
Via Bonardi 9, 20133 Milano, Italy
mailto: francesca.ieva at fastwebnet.it
francesca.ieva at mail.polimi.it
Voice: +39 02 2399 4604
Skype: francesca.ieva
-----
