Combining two distributions

Questions
6

So a quick follow-up on that, as expected it returns an asymmetric Student-T, which is good, but it’s not normalized to the peak:

import pymc3 as pm
import theano.tensor as tt

modeled = 0

observed = -0.001
sigma = tt.switch(tt.gt(observed, modeled), 1., 20.)
print(pm.StudentT.dist(2, mu=modeled, sigma=sigma).logp(observed).eval())

observed = 0.001
sigma = tt.switch(tt.gt(observed, modeled), 1., 20.)
print(pm.StudentT.dist(2, mu=modeled, sigma=sigma).logp(observed).eval())

Returns:
-4.0354533
-1.0397215

I would assume the distribution has to be continuous, but maybe pymc3 does something under the hood I don’t understand.