search

kgori / sigfit

Flexible Bayesian inference of mutational signatures
GNU General Public License v3.0
33 stars 8 forks source link

divergent transitions after warmup #67

Closed ab08028 closed 1 year ago

ab08028 commented 1 year ago

Hi there, I am using sigfit to extract mutation signatures from polymorpism data, and am running into the Stan error "There were X divergent transitions after warmup." I though this would be fixed by simply increasing the iterations, but have raised them up to 100,000 and am still getting the warning.

The warning occurs when I'm trying to determine the optimal 'k' number of signatures to extract (from 2 to 6). It is not clear to me when the warnings print whether they are for every model, or just the lower values of k.

Do I need to raise the iterations even higher? Is this an indication that the models with certain values of k don't fit well? Anything else you could suggest? Can these warnings be safely disregarded?

Thanks so much for your help! :)

ab08028 commented 1 year ago

Adjusting adapt_delta to 0.99 seems to have improved things! I think it's okay now. Sorry to bother you! :)