Hi Martin,
Thank you for your concrete answers. It turns out that I have falsly specify the y value.
y_observed = pm.Normal(“y_observed”, mu=f, sigma=sigma, observed=y[:,None])
it should be: y_observed = pm.Normal(“y_observed”, mu=f, sigma=sigma, observed=y)
Following with your suggestions, I am wondering if you have any suggestions for distributions to specify the prior. What is good priors for small lengthscales and what is good priors for large lengthscales? From my class, the teacher usually says a Gaussian/tructed Gaussian is good choice,but I get lost in so much prior distributions I can choose from in PYMC3.
Thank you very much!!!