number of parameters

questions concerning analysis/theory using programs M-SURGE, E-SURGE and U-CARE

number of parameters

Postby Sébastien » Tue Jan 09, 2007 4:50 pm

Hi,
I am using M-Surge 1.7.1 to estimate some survival and transition probabilities with 3 states. I have some issues with the number of parameters used to calculate AIC.
It seems that the rank of a particular model is a function of the previous model tested (that is, the same model will have a different rank function of the model tested just before). This does still occur when using random initial variables values.
Can anyone help me to understand (and fix!) this problem?
Thanks
Sebastian
Sébastien
 
Posts: 3
Joined: Fri Apr 28, 2006 2:17 pm
Location: UQAR

Postby CHOQUET » Mon Feb 05, 2007 4:46 am

Hello Sebastian,

First, have you try the last version, M-SURGE 1.8.1 ?.
Secondly, I don't really understand except that the rank seems to be dependent from the initial starting values. This kind of unstability can
occurs with models with a lot of parameters in comparison with the data.
In this kind of figure, the likelihood is close to flatness and ill conditioned for numerical problems. This can be one of the reason for this problem.
Is it the case for you ?

Rémi
CHOQUET
 
Posts: 211
Joined: Thu Nov 24, 2005 4:58 am
Location: CEFE, Montpellier, FRANCE.


Return to analysis help

Who is online

Users browsing this forum: No registered users and 7 guests

cron