Hi @cluhmann
Thanks for responding. That is exactly what I am trying to figure by learning about how sample_prior_predictive works. The way my problem is designed, I dont have the pdf of the model that I am using. Therefore I am not able to generate synthetic data from the process. My understanding is, you fix the parameters of the model, generate samples from the process, this becomes your ground truth. And then you run your MCMC sampler with observed=groundTruth, to see if the parameter samples are close to what you chose in the to generate the ground truth.
Basically my model is a cox proportional hazards model :
The baseline time to failure ~ LogNormal(mu,sigma) - if this was the only part in the model, I could have generated synthetic data(observed values) from this. But its the second multiplicative term due to which I am not able to generate synthetic data.
Trying to see how sample_prior_predictive can help in this situation