Fixing parameter values in HDDMRegressor

Reed Orchinik

unread,

Dec 26, 2022, 1:20:49 PM12/26/22

to hddm-users

Hi all,

I'm looking to do model comparison using HDDMRegressor between models that have parameters (specifically z and a) set to particular values. Let's say I want to set a = 1 and z = c, where c is a constant that I calculate for each observation (varies at the subject-round level). I also want to include "v ~ x" as a regression equation. Does anyone know a way to do this?

Thanks so much,

Reed

Reed Orchinik

unread,

Jan 9, 2023, 10:38:23 AM1/9/23

to hddm-users

Hi all,

Sorry to message again but just wanted to re-up this question. Is there a way to set a parameter to a numeric constant (which may change round to round) in HDDMRegressor? A post from 2012 suggested setting priors that have no variance. Is this still possible/a good way to approach the issue?

Thanks!

Reed

Yi'kang Liu

unread,

Jan 14, 2023, 2:08:32 AM1/14/23

to hddm-users

Hi Reed,

I am looking for the post from 2012, could you provide me the link for it?

Also, I am wondering if you mean that for group level the specific parameter has a fixed value, but it will vary for the subject level. I think it could be challenging to implement in the current API, as it may require modifying the raw code of the prior. It may also involve using a more complex hierarchical model.

Thanks,

Yikang

Yi'kang Liu

unread,

Jan 14, 2023, 2:13:14 AM1/14/23

to hddm-users

And you can refer to the post https://groups.google.com/g/hddm-users/c/PMXJO65FM6w/m/4Wb2xOhuAgAJ

Reed Orchinik

unread,

Jan 14, 2023, 12:00:07 PM1/14/23

to hddm-users

Hi Yikang,

Thanks for your response. It was actually two posts, the first about fixing parameters in the LAN extension: https://groups.google.com/g/hddm-users/c/i2NgmrDDTg4/m/nSeTQmCdBgAJ; and the second about setting priors in an older version of HDDM: https://groups.google.com/g/hddm-users/c/fuFMy_35zfI/m/qhJz4iOfAgAJ

It seems like the post you link to has more recent information than the second post I reference. Thanks!

I'm actually thinking of fixing the parameter (bias) at the individual level. Currently, I'm estimating the model separately for every person in the data set because we have specific predictions about the information that an individual is using and want to test various models against each other at the individual-level. Does that seem do-able?

Thanks again,

Reed

Yi'kang Liu

unread,

Jan 21, 2023, 4:45:05 AM1/21/23

to hddm-users

Hi Reed,

HDDM has updated to 0.9.8, and in this version, maybe we can set the prior. You can refer to https://hddm.readthedocs.io/en/latest/tutorial_set_parameter_defaults.html.