Changing number of states does not change model fitting

abuzarmahmood / pytau

Library to streamline batch-fitting of Bayesian changepoint models to neural population spiking data

https://abuzarmahmood.github.io/pytau/

MIT License

1 stars 0 forks source link

Changing number of states does not change model fitting #34

Closed cmazzio closed 4 weeks ago

cmazzio commented 1 month ago

Changing states to 3 produces model fitting with 4 states:

Screenshot from 2024-06-02 13-37-21

abuzarmahmood commented 1 month ago

Could you please add where you changed number of states and what code you ran to fit the models and generate the plots?

cmazzio commented 4 weeks ago

I rat fit_with_fit_handler.py and changed only the states variable to 3

abuzarmahmood commented 4 weeks ago

Could you please attach the updated scripts you used here. Thanks.

abuzarmahmood commented 4 weeks ago

Issue selecting correct model

https://github.com/abuzarmahmood/pytau/blob/d0a2d6e697579bc934ae802a26a41fd2fa059e33/pytau/how_to/scripts/fit_with_fit_handler.py#L57C1-L61C49

dframe = fit_database.fit_database
wanted_exp_name = 'pytau_test'
wanted_frame = dframe.loc[dframe['exp.exp_name'] == wanted_exp_name] 
# Pull out a single data_directory
pkl_path = wanted_frame['exp.save_path'].iloc[0]

Plots came out fine once correct model was selected.