I have a very basic inference problem, that has a single prior belief and a set of observed values that i want to infer a posterior from.
The thing here is that the values are extremely small. I have the following model:
obs = np.ones(100)*10**-6 prior_mean = 1e-8 prior_sd = 1e-6 with pm.Model() as model: prior = pm.Normal('prior', mu=prior_mean, sd = prior_sd) pm.Normal('likelihood', mu=prior, observed=obs) trace = pm.sample() pm.tracpelot(trace)
However, although the observed values are 100 times bigger than the prior, the posterior looks identical to the prior. This would not have happened if the prior was 0 and the observed values were 100 for example.
How do i deal with this small values issue? I have tried centering the prior without any result
Thank you for helping out!