#806 incorrectly removed zero-padding for NegativeBinomial2 distributions, on the incorrect assumption that the addition of support for zero-mean Poisson would resolve issues with zero-mean negative binomials. This was not the case. Our tests did not catch this. We should
- Add tests for NegativeBinomialNoise and NegativeBinomialObservation with
0 mean and confirm they currently fail.
- Add back epsilon padding to NegativeBinomialNoise and NegativeBinomialObservation
- Confirm the tests now pass.
#806 incorrectly removed zero-padding for
NegativeBinomial2distributions, on the incorrect assumption that the addition of support for zero-mean Poisson would resolve issues with zero-mean negative binomials. This was not the case. Our tests did not catch this. We should0mean and confirm they currently fail.