Changing priors

[Pat Simen]

Hi again,

I see that you can set "informative" to On or Off. But I was curious whether there was an easy way to get in and change the parameter values of HDDM's priors.  I'm thinking of, e.g., changing the mean of the Ter distribution prior to 0.25.  Is that easy to do? 

Also, is it correct that if the data really support Ter=0.25, then the effect of a prior that Ter is most likely to be 0.4 will be wiped out by sufficient data? I assume so, but just wanted to check.

Thanks again -- HDDM seems to be working quite well!  

--Pat

[Thomas Wiecki]

Hi Pat,

There's no parameter for that. But it's quite easy to build your own model that does that. For that, inherit from HDDM.HDDM, and overwrite the _create_stochastic_knodes_info() method with a copy from here: https://github.com/hddm-devs/hddm/blob/master/hddm/models/hddm_info.py#L128 and change the priors as you see fit.

But yes, since these are group priors with a few subjects the data should overwhelm it. Do note that the parameter recovery studies showed that even with little data you get good estimates with true parameters away from the priors.

Thomas

--
You received this message because you are subscribed to the Google Groups "hddm-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hddm-users+...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

--

Thomas Wiecki, PhD

Data Science Lead, Quantopian Inc, Boston

[hcp...@gmail.com]

Hi, Thomas,

This is an old post, but just wondering, did you have specific papers in mind when you said "the parameter recovery studies showed that even with little data you get good estimates with true parameters away from the priors". I'd like to dig a bit on this. Thanks in advance.

Best,

Chuan-Peng