Please see if you know the answers to four questions concerning the problems described below. I followed @bgoodri ’s suggestion in this post to turn centered parametrization to non-centered parametrization of correlated parameters. This significantly improve the accuracy of the parameter estimates b…

Thanks for the clarification. I successfully stan_rdump() the data. In installing cmdstan, I successfully completed the make build step. However, when compiling the .stan file (which I have been using with rstan without a problem), I received the errors below. In another (slower) cluster, I could co…

Also, I followed p.28 of the cmdstan-guide-2.19.1.pdf to hope to sample 4 chains in parallel but the chains ran sequentially even when multiple cores are specified. Any suggestion or pointer to the correct commands for a cluster environment? #PBS -l nodes=1:ppn=8 for i in {1..4} do ../SSM0.5.2dev…

Did you get things to compile? I’m guessing this is the compiler. I think Stan supports gcc 4.9.3 (whatever Rtools is at). std::index_sequence is a c++14 thing ( https://en.cppreference.com/w/cpp/utility/integer_sequence ) To do four chains in parallel, you gotta run four separate things. If I rememb…

Hi, Ben: I finally completed a run of 4 chains using metric=dense_e (to compare with that using metric=diag_e). It does reduce the run time significantly (with essentially the same estimates given the large amount of simulated data used: 12,000 data points). Here are the summary output: dense_e_cm…

Well the things to watch out for are the usual things: are there divergences in the dense_e version and are the Rhats 1? Your N_effs still seem really high. You might be able to get away with cutting your chain lengths in half or something. Maybe try 500 warmup/500 samples and see what happens? I w…

[image] bbbales2: are there divergences in the dense_e version and are the Rhats 1? Would appreciate your clarification of the following issues that seem confusing. In the documentation (eg, p.29 of cmdstan-guide-2.19.1.pdf ), the reported divergent__ row has the NEff equal to the 1000 iterna…

Don’t worry about Rhat for the divergent__ parameter, that’s for the parameters mostly. 5e-3 is the fraction of post-warmup trajectories that were divergent in all chains, so there were 5e-3 * 1600 = 8 divergences across all four chains. This model looks like it took way longer than the previous o…

[image] bbbales2: Rhat occasionally ends up less than 1.0. I don’t think what you’re seeing in your results here it’s a thing to worry about. I have been assuming that the 1.01 threshold for Rhat you mentioned before is equivalent to a 0.99 threshold. So only the deviation from 1 matters, rig…

[image] tlyim: So only the deviation from 1 matters, right? I think Rhat is supposed to be greater than or equal to 1.0. I think it being less than one just happens randomly, but isn’t a big deal. So less than or greater than one are interpreted differently. [image] tlyim: I suspect that…

Thanks for your clarification and suggestion. Make a lot of sense. Owing to some other considerations, for now I still use 10,000 simulated data points. I am returning to more conservative choices of warmup (1000) and delta (0.99) to first ensure no divergence. Just finished a comparison of the den…

Problems remain after non-centered parametrization of correlated parameters

Modeling

tlyim June 25, 2019, 3:23pm 32

Thanks for the tips. I summarized the current status of my exploration in this post.

Topic		Replies	Views
Adaptation and non-centered parameterisations Developers	10	1283	August 2, 2021
Divergences in a non-centered computational model Modeling fitting-issues	21	1558	October 30, 2019
How to speed up sampling in rstan? Modeling rstan , performance , hierarchical-model	6	6259	July 24, 2020
How to determine the combination of adapt_delta / max_treedepth / metric=dense_e vs. diag_e / term_buffer / window / warmup / iter? Modeling	2	1710	May 8, 2019
Non-centered parameterization for non-hierarchical parameters? Modeling reparametrization	15	1534	May 6, 2021

Problems remain after non-centered parametrization of correlated parameters

Related topics