Unknown bug/problem when running a hierarchical HSSM (Hierarchical=True)

[noah3695]

Hi HSSM community,

I have a dataset where a total of 26 participants completed a task and I have roughly 47,000 trials across all participants.

I have run a non-hierarchical model with some success. The non-hierarchical model was made using:
simple_ddm_model3 = hssm.HSSM(data=data, include=[ { "name": "v", "formula":"v ~ 1 + (1|blurVal)", } ],)

I've also attached two figures: model fit and traces of the posteriors. The model seems to fit reasonably well.

.

I ran 16 chains, 8000 samples, and tune-in set to 1000. simple_ddm_model3, when probed, suggests it is a hierarchical model:

Hierarchical Sequential Sampling Model
Model: ddm

Response variable: rt,response
Likelihood: analytical
Observations: 10457
...

However, I remember there being an argument to creating the model where you can set hierarchical=True. Examples of this can be found in multiple other submissions from other members, including, here, here and here.

---

When I try to run the model with hierarchical=True, i get the following error(s)

Traceback (most recent call last):

File ~/anaconda3/envs/HSSM/lib/python3.11/site-packages/spyder_kernels/py3compat.py:356 in compat_exec
exec(code, globals, locals)

File ~/Documents/HDDM_modelfit_v1.py:445
HDDM_modelfit_v1(sampler_input, cores_input, chains_input, draws_input, tune_input)

File ~/Documents/HDDM_modelfit_v1.py:187 in HDDM_modelfit_v1
simple_ddm_model3 = hssm.HSSM(data=cav_data,

File ~/anaconda3/envs/HSSM/lib/python3.11/site-packages/hssm/hssm.py:296 in init
self._override_defaults()

File ~/anaconda3/envs/HSSM/lib/python3.11/site-packages/hssm/hssm.py:1010 in _override_defaults
param_obj.override_default_priors_ddm(

File ~/anaconda3/envs/HSSM/lib/python3.11/site-packages/hssm/param.py:214 in override_default_priors_ddm
dm = self._get_design_matrices(data, eval_env)

File ~/anaconda3/envs/HSSM/lib/python3.11/site-packages/hssm/param.py:270 in _get_design_matrices
dm = design_matrices(formula, data=data, extra_namespace=extra_namespace)

File ~/anaconda3/envs/HSSM/lib/python3.11/site-packages/formulae/matrices.py:556 in design_matrices
design = DesignMatrices(description, data, env)

File ~/anaconda3/envs/HSSM/lib/python3.11/site-packages/formulae/matrices.py:58 in init
self.response.evaluate(data, env)

File ~/anaconda3/envs/HSSM/lib/python3.11/site-packages/formulae/matrices.py:137 in evaluate
self.term.set_type(self.data, self.env)

File ~/anaconda3/envs/HSSM/lib/python3.11/site-packages/formulae/terms/terms.py:820 in set_type
self.term.set_type(data, env)

File ~/anaconda3/envs/HSSM/lib/python3.11/site-packages/formulae/terms/terms.py:431 in set_type
component.set_type(data)

File ~/anaconda3/envs/HSSM/lib/python3.11/site-packages/formulae/terms/variable.py:90 in set_type
raise ValueError(f"Variable is of an unrecognized type ({type(x)}).")

ValueError: Variable is of an unrecognized type (<class 'pandas.core.series.Series'>).

Traceback (most recent call last):

Cell In[15], line 1
simple3, simple_ddm_model3 = runfile('/Users/steinbergnj/Documents/HDDM_modelfit_v1.py', args='nuts_blackjax 8 8 5000 1000')

TypeError: cannot unpack non-iterable NoneType object

---

Any help to verify if it's already somehow hierarchical, or how to get it to run hierarchically using the input hierarchical=True would be much appreciated. Thanks!

[frankmj]

The way you have modeled it here treats blurval hierarchically but not participants. It is not recommended to use task condition (assuming that is what blurval is) in a hierarchical way unless you have at least 5 conditions -- otherwise there is not enough data to inform the sigma (see issue #528). Instead you can do v ~ 1+ blurval (and then add hierarchy across subjects later if you want - to do that you would do v ~ 1 + blurval + (1+blurval|participant_id)). Setting hierarchical = true only applies to the intercepts and so in general it is better to fully specify your hierarchical model yourself Michael

…

On Wed, Sep 4, 2024 at 10:49 AM noah3695 ***@***.***> wrote: Hi HSSM community, I have a dataset where a total of 26 participants completed a task and I have roughly 47,000 trials across all participants. I have run a non-hierarchical model with some success. The non-hierarchical model was made using: simple_ddm_model3 = hssm.HSSM(data=data, include=[ { "name": "v", "formula":"v ~ 1 + (1|blurVal)", } ],) I've also attached two figures: model fit and traces of the posteriors. The model seems to fit reasonably well. modelFit_allSubj_VPrior.png (view on web) <https://github.com/user-attachments/assets/13287e90-2e73-483b-8b44-1bdab6764482> plotTracerace_log_likelihood_samplernuts_blackjax_cores8_chains16_draws8000_tune1000.png (view on web) <https://github.com/user-attachments/assets/41fe8ffa-d2a1-4294-b64d-10b85ad0e3ea> . I ran 16 chains, 8000 samples, and tune-in set to 1000. simple_ddm_model3, when probed, suggests it is a hierarchical model: Hierarchical Sequential Sampling Model Model: ddm Response variable: rt,response Likelihood: analytical Observations: 10457 ... However, I remember there being an argument to creating the model where you can set hierarchical=True. Examples of this can be found in multiple other submissions from other members, including, here <#306>, here <#304> and here <#565>. ------------------------------ When I try to run the model with hierarchical=True, i get the following error(s) Traceback (most recent call last): File ~/anaconda3/envs/HSSM/lib/python3.11/site-packages/spyder_kernels/py3compat.py:356 in compat_exec exec(code, globals, locals) File ~/Documents/HDDM_modelfit_v1.py:445 HDDM_modelfit_v1(sampler_input, cores_input, chains_input, draws_input, tune_input) File ~/Documents/HDDM_modelfit_v1.py:187 in HDDM_modelfit_v1 simple_ddm_model3 = hssm.HSSM(data=cav_data, File ~/anaconda3/envs/HSSM/lib/python3.11/site-packages/hssm/hssm.py:296 in *init* self._override_defaults() File ~/anaconda3/envs/HSSM/lib/python3.11/site-packages/hssm/hssm.py:1010 in _override_defaults param_obj.override_default_priors_ddm( File ~/anaconda3/envs/HSSM/lib/python3.11/site-packages/hssm/param.py:214 in override_default_priors_ddm dm = self._get_design_matrices(data, eval_env) File ~/anaconda3/envs/HSSM/lib/python3.11/site-packages/hssm/param.py:270 in _get_design_matrices dm = design_matrices(formula, data=data, extra_namespace=extra_namespace) File ~/anaconda3/envs/HSSM/lib/python3.11/site-packages/formulae/matrices.py:556 in design_matrices design = DesignMatrices(description, data, env) File ~/anaconda3/envs/HSSM/lib/python3.11/site-packages/formulae/matrices.py:58 in *init* self.response.evaluate(data, env) File ~/anaconda3/envs/HSSM/lib/python3.11/site-packages/formulae/matrices.py:137 in evaluate self.term.set_type(self.data, self.env) File ~/anaconda3/envs/HSSM/lib/python3.11/site-packages/formulae/terms/terms.py:820 in set_type self.term.set_type(data, env) File ~/anaconda3/envs/HSSM/lib/python3.11/site-packages/formulae/terms/terms.py:431 in set_type component.set_type(data) File ~/anaconda3/envs/HSSM/lib/python3.11/site-packages/formulae/terms/variable.py:90 in set_type raise ValueError(f"Variable is of an unrecognized type ({type(x)}).") ValueError: Variable is of an unrecognized type (<class 'pandas.core.series.Series'>). Traceback (most recent call last): Cell In[15], line 1 simple3, simple_ddm_model3 = runfile('/Users/steinbergnj/Documents/HDDM_modelfit_v1.py', args='nuts_blackjax 8 8 5000 1000') TypeError: cannot unpack non-iterable NoneType object ------------------------------ Any help to verify if it's already somehow hierarchical, or how to get it to run hierarchically using the input *hierarchical=True* would be much appreciated. Thanks! — Reply to this email directly, view it on GitHub <#580>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAG7TFFQ4FKVZIZBV5LYUXDZU4MXPAVCNFSM6AAAAABNUQCWAGVHI2DSMVQWIX3LMV43ERDJONRXK43TNFXW4OZXGEZTOOJYGU> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>