The Hurdle Gamma likelihood does not work with with even the native sampler. I have been using an earlier of version of BAMBI and PYMC for a while, when the Hurdle Gamma likelihood worked (bambi ==0.15.0 || pymc==5.23.0 || pytensor==2.31.4).
I am currently using:
bambi == 0.16.0
pymc==5.27.0
pytensor ==2.36.1
Here is example code:
import bambi as bmb
import polars as pl
import pymc as pm
# create simple dataframe with HG likelihood
dist = pm.HurdleGamma.dist(psi=0.3, mu=4, sigma=0.5)
draws = pm.draw(vars=dist, draws=100, random_seed=1)
df = pl.DataFrame({"area": draws}).to_pandas()
# create bambi model
model = bmb.Model(formula="area ~ 1", data=df, family="hurdle_gamma")
# sample
idata = model.fit()
idata
I get:
The Hurdle Gamma likelihood does not work with with even the native sampler. I have been using an earlier of version of BAMBI and PYMC for a while, when the Hurdle Gamma likelihood worked (bambi ==0.15.0 || pymc==5.23.0 || pytensor==2.31.4).
I am currently using:
bambi == 0.16.0
pymc==5.27.0
pytensor ==2.36.1
Here is example code:
I get: