How to make truncated Student-T?

As per title. How do I build a truncated Student-T likelihood?

You can use pm.Bound if you dont mind the log_prob being not normalized, otherwise you need to implement one yourself like https://github.com/pymc-devs/pymc3/blob/1b836c67cb7a56662f5c33e3a3aec72ecd683592/pymc3/distributions/continuous.py#L563 which subtract the cdf outside of each bound.

1 Like