In my data set is sometimes happens that regardless of how large I set the maximum iterations (I stopped when 150000 was not sufficient), the maximum iterations is reached. When I ran the algorithm ten times, only one time I ended up with a hit on that maximum but this run was the run with the best values and thus presented to me. From the LnL Development report it all looks normal for the first few hundred iterations but then the delta starts to switch from negative to positive randomly.
Is this normal behavior that can happen with moderately correlated data? If necessary I can provide the final report.
Thanks for your help!
Jan
FIMIX algorithm not converging
-
- PLS Expert User
- Posts: 116
- Joined: Fri Sep 12, 2014 2:12 pm
- Real name and title: Jan Schreier
-
- SmartPLS Developer
- Posts: 1284
- Joined: Tue Mar 28, 2006 11:09 am
- Real name and title: Dr. Jan-Michael Becker
Re: FIMIX algorithm not converging
You should not use those results from a run that did not converge correctly. The process sometimes does that and jumps between two local optima. Usually those results are conceptually inferior to other converged solutions.
Dr. Jan-Michael Becker, BI Norwegian Business School, SmartPLS Developer
Researchgate: https://www.researchgate.net/profile/Jan_Michael_Becker
GoogleScholar: http://scholar.google.de/citations?user ... AAAJ&hl=de
Researchgate: https://www.researchgate.net/profile/Jan_Michael_Becker
GoogleScholar: http://scholar.google.de/citations?user ... AAAJ&hl=de
-
- PLS Expert User
- Posts: 116
- Joined: Fri Sep 12, 2014 2:12 pm
- Real name and title: Jan Schreier
Re: FIMIX algorithm not converging
Thank you for your help, Michael!
-
- PLS Expert User
- Posts: 116
- Joined: Fri Sep 12, 2014 2:12 pm
- Real name and title: Jan Schreier
Re: FIMIX algorithm not converging
Dear Jan,
sorry for reviving this thread. In my case the lnL-Values are often way better if I get stuck in a local optimum (e. g. lnL=8.000) compared to those cases where the algorithm converts as expected (e. g. lnL= -20.000). Would you still discard the local optimum solution?
Thanks for your help!
Jan
sorry for reviving this thread. In my case the lnL-Values are often way better if I get stuck in a local optimum (e. g. lnL=8.000) compared to those cases where the algorithm converts as expected (e. g. lnL= -20.000). Would you still discard the local optimum solution?
Thanks for your help!
Jan