Prooblem about Model and Distribution

Dear Stan user:
I would like to know whether my stan model and distribution are fitted properly.
Is there any index I could check(like n_eff,Rhat…) without thousands of iterations?(because many of iterations take a long time)

Any response could help,thanks!

Hi Eric,

have you tried to fit your model using only one chain and 1000 iterations? You can do that as a first step and see what happens.

1 Like

I use pystan.And I would like to predict the stiffness of tumor by “pressing depth”,“force curve” and “tumor morphology”. The image below is the model I assign.

I have try to fit my model by 1000 iterations.This is the result I get.How could I know whether I have a proper model and distribution or not?
results

try putting priors on your parameters -
https://github.com/stan-dev/stan/wiki/Prior-Choice-Recommendations

the basics and explanation for PyStan are here:
https://mc-stan.org/users/documentation/case-studies/pystan_workflow.html

the other suggestion is to choose what you think are reasonable values for your parameters, generate some data using those values, fit the model to the simulated data, and check the estimates for the parameters. more details here:

1 Like

Thanks for your detailed information!

One more question,after finishing all of this chek(Tree Depth, E-BFMI).Only the divergence is not small enough.
Do I have to restart iterate again(Like the following link)?

Get intermediate results in pystan

divergences indicate a problem - you need to fix the model.
did you add priors?

I don’t know about pystan.
if this is taking a long time, you should create simulated data and use that to test the model.

I’ll keep in mind and try.
Thanks for your useful informations!