I think the difference between the two cases, is the clip introduces a discontinuity / flat region in the gradient but does not avoid the sampler for seeing it, whereas your original uniform completely avoids the impossible region.
Related topics
| Topic | Replies | Views | Activity | |
|---|---|---|---|---|
| Define constraint on the priors | 12 | 2147 | April 19, 2018 | |
| Sampling with a general lognormal model | 9 | 1004 | December 3, 2019 | |
| How do I implement an upper limit log normal distribution | 6 | 2499 | June 8, 2018 | |
| Truncated Inverse normal distribution (also known as Wald distribution) | 15 | 2955 | April 27, 2018 | |
| Convergence issues with LogNormal | 2 | 546 | March 16, 2022 |