convergence problems with a large number of sessions

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

convergence problems with a large number of sessions

Postby vincenzogervasi » Thu Sep 11, 2014 6:21 am

Hi,
I am trying to run a multi-season occupancy model with a large number of sessions (64 each year for 6 years, summing up to 384 session). When I run the model, the deviance starts descending as expected, but then jumps up to a very large number (8.21841e+307) and never changes again, showing some clear convergence problem. While I am not surprised, given the complexity of my dataset, do you have any advice on how to try and improve convergence, maybe by changing some of the advanced numerical options in E-SURGE?

Thanks a lot.
vincenzogervasi
 
Posts: 3
Joined: Tue Jul 29, 2014 5:57 am

Re: convergence problems with a large number of sessions

Postby Guillaume Souchay » Mon Oct 27, 2014 10:59 am

Hi Vincenzo

I don't know if you still have some troubles with your models.
To improve the convergence, you may choose your own initial value for beta if you have some ideas of the values for your parameter.
What options are you using by now? Maybe the algorithm Em(20) + QN can be useful, especially if you have boundary parameters.
Multiple cycles can be useful (the "stop after 1 cycle" box), sometimes, it can "break" to infinite deviance trend: when the numbers of iteration is exceeded, it usually restarts from the last good deviance point ...

Good luck with your huge model !

Cheers,
Guillaume
Guillaume Souchay
 
Posts: 64
Joined: Wed Feb 27, 2008 5:20 am
Location: Nantes, FRANCE


Return to analysis help

Who is online

Users browsing this forum: Google [Bot] and 8 guests