search

Genentech / jmpost

https://genentech.github.io/jmpost/
17 stars 4 forks source link

Inter-indiv variability on phi in the GSF model #283

Closed mercifr1 closed 8 months ago

mercifr1 commented 8 months ago

@gowerc, If I interpret this correctly, in the GSF model, the phi parameter has no inter-individual variability (IIV). right? Do you think we could introduce IIV taking advantage of an inv_logit() transformation, like this:

  vector[Nind] psi_phi =  inv_logit(mu_phi[arm_index] + eta_tilde_phi * omega_phi);

Ref. to: here.

gowerc commented 8 months ago

By inter-individual variability do you mean per patient effect? If so then I think that is covered by this line:

    // parameters block
    vector<lower={{ machine_double_eps }}, upper={{ 1 - machine_double_eps }}>[Nind] lm_gsf_psi_phi;
    vector<lower={{ machine_double_eps }}>[n_arms] lm_gsf_a_phi;
    vector<lower={{ machine_double_eps }}>[n_arms] lm_gsf_b_phi;

    // model block
    lm_gsf_psi_phi ~ beta(lm_gsf_a_phi[pt_arm_index], lm_gsf_b_phi[pt_arm_index]);

That is each subject gets their own individual Phi parameter and that the priors for the patient level Phi parameter are set based upon which treatment arm they belong to.

mercifr1 commented 8 months ago

Yep, I see. I overlooked this. Sorry about that. Thanks a lot Craig.