Seeking for suggestion on HDDM for PRT task

[徐钰佳]

Dear all,

We conducted a study using PRT task (Bogdan&Pizzagalli, 2006) , which invetigates the reward learning under non-stress and stress condition respectively. Currently we are applying HDDM to compare group difference, but meet a question: we can set up non-stress HDDM and stress HDDM for five groups respectively, and got results, but non-stress and stress is within-subject, we are confused how to set up a HDDM, including a 2-level within-subject variable and 5 groups(2*5, Repeated Measures ANOVA), could you please give some variable thoughts.

1) How to add the within-subject variable(nonstress, stress) to one HDDM model given that we can’t just use “depends_on” for it? We've tried HDDMRegressor, but it seemed inappropriate(bad convergence, high DIC and no reflection of interaction).

The code for HDDMRegressor is below:    

m_reg = hddm.HDDMRegressor(data, "v ~ C(Stress, Treatment(‘NoStr’))",

depends_on = {'v': 'group','a': 'group', t':'group','z':'group'},

include = ['v', 'a', 't', 'z'])

m_reg.find_starting_values()

m_reg.sample(10000, burn = 1000, dbname = 'All_Reg_vatz.db', db = 'pickle')

m_reg.save('All_Reg_vatz')

2) We tried to divide 10 groups(MDD_Nostr, MDD_Str, BPD_Nostr, BPD_Str ...) to fit the model but how to compare the parameters between groups? Can I use Bayesian Repeated Measure ANOVA?

Many thanks & Apology for my naive questions

Best Wishes