Having problems defining random function for custom distribution with mixture of observed and random variables

Hmmm, maybe an easier way to do is to extract the logalpha and beta from the posterior (i.e., trace), and build the posterior prediction logalpha + beta * time, and logistic transformed it and generate Bernoulli random?