search

venpopov / bmm

An R package for easy and flexible Bayesian Measurement Modeling
https://venpopov.github.io/bmm/
GNU General Public License v2.0
12 stars 3 forks source link

[bug] `sd` for `thetant` for setsize = 1 not fixed to zero in `mixture3p` #202

Closed GidonFrischkorn closed 6 months ago

GidonFrischkorn commented 6 months ago

Describe the bug When working on the revisions of the tutorial paper I noticed that when we estimate the mixture3p using random effects for the set_size variable that there are still random effects estimated for thetant for set size = 1. See screenshot below

image

To Reproduce You can see the specified call to bmm in the screenshot of the summary.

Expected behavior Ideally, we should fix the sd to zero and list them under Constant Parameters. But I do not know if this is possible via priors. There might be constraint that sd parameters cannot be fixed to zero, but only a value very close to zero.

GidonFrischkorn commented 6 months ago

I checked this again, and it actually is not a serious problem, as all non-target mixtures are switched off using the Index-Coding we implemented. So, likely the estimated sd in these model just represents samples from the prior. The same happens in the imm models for the a and s parameters for set size = 1

venpopov commented 6 months ago

Yes, I noticed that some time ago but concluded the same that it's not a problem because we turn the effects off. It could still be confusing to users, so I'll see if there's an easy way to fix those to a constant too

venpopov commented 6 months ago

it was an easy solve, PR is up