Enrique Cardenas III

Results 9 comments of Enrique Cardenas III

@lbittarello, Thank you for your response, it has been very helpful. Distributional assumptions would be key here, since I'm modeling on Tweedie distributed response data. In that case, I have...

@lbittarello, Thank you for your expeditious reply. To the first part of your reply, I think I understand. It seems like you are saying that a manual grid search won't...

@lbittarello, This makes a lot of sense looking back. Giving weights of 1 for all records proves to give equal deviance results. Alternatively, using a manual [family.deviance](https://www.statsmodels.org/dev/generated/statsmodels.genmod.families.family.Family.deviance.html#statsmodels.genmod.families.family.Family.deviance) calculation with mu...

@lbittarello, Can it be true that the predictions are the same? If so, what would differentiate `glum`'s elastic net from `statsmodels` without elastic net? Using the predict functions for each...

@lbittarello, Got it. So in exchange for the chance of a bump in deviance on the train dataset, the elastic net models are expected to generalize better to unseen data....

> For example, you could cross-validate the metric of interest (the deviance, in this case) to have an idea of what it looks like on unseen data. What do you...

Think I've got a draft for a solution that strips some of the statsmodels code down. I believe I'll only be working with models that have identity link, but I'll...

Not sure if the above code is entirely correct, actually. It's a good starting point, but the statsmodels method is a bit more complicated than I originally thought. Are there...

Here's something I've been working on, based on statsmodels v0.13.5. It's missing a few features present in the statsmodels `generalized_linear_model.py` source code for the [`get_prediction`](https://www.statsmodels.org/dev/_modules/statsmodels/genmod/generalized_linear_model.html#GLMResults.get_prediction) function and its subfunctions (Prediction...