I have the following basic model where theta (\theta) is a measured value that may predict whether someone reports tinnitus and has_tinnitus is a binary array indicating whether or not the particular subject has tinnitus:
We have some pretty good prior knowledge about the prevalence of tinnitus in the general population. However, it can be pretty challenging to transform this information into priors for \beta_0 and \beta_1. But, we can easily specify \theta_i \sim Beta(\alpha_i, \beta_i). According to Equation 2 in Bedrick et al. 1996 (A new perspective on priors for generalized linear models; http://www.jstor.org/stable/2291571), we can come up with priors for two different values of \theta_i and then place an induced prior on . For example, if I have \theta_1 \sim Beta(\alpha_1, \beta_1) and \theta_2 \sim Beta(\alpha_2, \beta_2), then the joint prior for \delta (i.e., d0 and d1 in the model) is
Iām a bit stumped about how to implement this. I tried searching for joint prior and conditional means prior, but could not find any information about either on the PyMC3 forums. Any advice?