-
- [x] Initializers for model parameters
- [x] Splines for baseline hazards
- [x] Fix ISplines for baseline intensities - constrain all parameters to be the same sign, currently constrained to all be…
-
On many datasets I notice warnings like:
`WARNING: Log-likelihood -7226.94 of K2P+R4 worse than K2P+R3 -7059.91`
Obviously this shouldn't happen. The lnL should always be better with R4 than R3.…
-
Need to prepare package for rOpensci
- [x] install `autotest` and run it on package
- [ ] run autotest on Tidydensity
-
```
venture[script] > observe (normal 0 1) 10
venture[script] > global_log_likelihood
[0]
```
This is because global_log_likelihood is implemented in terms of regen, and regen only ever cares about t…
-
When processing a dataset spanning more than 6 months, the program crashes with a memory allocation error at the step "Training end evaluating model."
An error like this:
"Traceback (most recent c…
-
Using TreeDataLikelihood
Branch rate model used: strictClockBranchRates
Using BEAGLE DataLikelihood DelegAug 15, 2024 6:05:44 PM dr.app.beast.BeastMain
SEVERE: Parsing error - poorly formed BEAS…
-
looks like this is actually pretty easy to add this generically.
We need to be able to build the unrestricted model with access to score_obs and hessian evaluated at some params (already available …
-
By implementing a linear-link, positive basis & positive weight (either constrained optimization or convexity preserving parametrization) version of the GLM, we can use the log-likelihood of each ad…
-
#### Describe the bug
I ran into these results when running a constrained optimization GLM using Tweedie(var_power=1.5). I believe more than the constrained optimization itself, this is related to …
-
### Is your feature request related to a problem? Please describe.
The task of finding the MLE for a three-parameter (generalized) Pareto distribution boils down to finding _all_ roots of the (univ…