-
**update** as FAQ
warnings about links that do not impose domain restrictions for the family
related FAQ issue #2208 which is mostly a discussion when we figured this out the first time.
my comment…
-
Is there any way to create an empty GLM Binomial Model? And then proceed to set the coefficients manually?
-
coefplot.glm produces graph when not specifying predictors but adding it produces:
`Error in `$
-
(intended as umbrella issue)
We need the diagnostics and specification tests for all models, similar to stats.dignostics, stats.influence_outliers and graphics.regressionplots which is currently main…
-
(random idea for evaluating how well a distribution matches the data)
In GLM models like tweedie or gamma or Poisson, the error is not additive as in a gaussian model (linear or nonlinear least squ…
-
IRLS produces results with no warnings when the fitted means are outside the family domain. It's not clear that the solution is actually an MLE or if the standard errors, etc. are meaningful. An exa…
-
The gamma GLM seems to fail easily during optimization when using the canonical reciprocal link. It works OK with clean simulated data, but fails with real data sets or complicated models. With the …
-
experimental idea
I think we can reuse large parts of GLM with IRLS with just estimating equations defined by mean/link and variance function.
This is similar to Quasi/Pseudo MLE in R but without de…
-
Now vikos is able to fit arbitrary glm models.
It would be nice if one adds the link functions for custom glm models such as Normal, Exponential, and so on.
You can find an overview here: [Table cus…
-
Why does GLM negative binomial default to pearson estimate of overdispersion?
unit tests for GLM negative binomial use `Committee` which has reference number scale=0.385...
scaletype is set in `fit`…