You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
We are interested in implementing the bas.glm() function cannot be applied in a quasi-Poisson or negative binomial regression model framework. It would be very useful to deal with over-dispersed count data. We look forward to this extension in the near future.
The text was updated successfully, but these errors were encountered:
planning on a version for Negative Binomial regression early 2023. Let me know if you are still interested or in want to be involved in testing.
Thank you very much, Merlise. We are not currently estimating this type of models in our research. However, we can help with checking the functioning of new versions with our data.
We are interested in implementing the bas.glm() function cannot be applied in a quasi-Poisson or negative binomial regression model framework. It would be very useful to deal with over-dispersed count data. We look forward to this extension in the near future.
The text was updated successfully, but these errors were encountered: