Exponential likelihood at different location

Questions
8

I couldn’t make the suggested shiftedExp code work, pymc has changed, but I found a CustomDist example which did work.

Just for the record:

– Malcolm

# Defined: https://www.pymc.io/projects/docs/en/v5.10.4/api/distributions/generated/pymc.CustomDist.html

from pytensor.tensor import TensorVariable

def shiftedExp(
    lam: TensorVariable,
    shift: TensorVariable,
    size: TensorVariable,
) -> TensorVariable:
    return pm.Exponential.dist(lam, size=size) + shift

with pm.Model() as linreg:
    lam = pm.HalfCauchy('sigma', beta=10, initval=1.)
    intercept = pm.Normal('Intercept', 0, sigma=20)
    x_coeff = pm.Normal('Slope', 0, sigma=20) 

    likelihood = pm.CustomDist('y', lam, intercept + x_coeff * x, observed=y, dist=shiftedExp)
    trace2 = pm.sample()
2 Likes