Different approaches to the Challenger Example

[junpenglao]

@einarbmag

I.e., my data is a sequence of tuples (x,t) where x is 0 or 1, and t is a positive number. The probability p of X to be 1 depends on the value t in some way.

I think this is usually modelled using a Poisson point process. I have never fit one myself, but using the GP module you can do something similar: http://docs.pymc.io/notebooks/GP-slice-sampling.html#Gaussian-Process-Classification