Timeline for Error metrics for cross-validating Poisson models
Current License: CC BY-SA 4.0
22 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Jun 11, 2019 at 22:23 | comment | added | Tom Wenseleers | @Glen_b Given that Poisson GLMs are fit using iteratively reweighted LS, could one also not use the R2 calculated from the weighted MSE of the weighted LS fit of the last IRLS iteration with which the GLM is fit, as in stats.stackexchange.com/questions/412580/… ? | |
| Oct 26, 2018 at 14:43 | history | edited | amoeba | CC BY-SA 4.0 |
edited tags
|
| Oct 5, 2013 at 18:47 | vote | accept | Zach | ||
| Oct 5, 2013 at 18:46 | comment | added | Zach | Ok, thanks for all the comments-- they've been very useful. | |
| Oct 4, 2013 at 23:17 | comment | added | Glen_b | Thanks for responding, Zach. In that case, I don't know. I think that it's still possible to generate an analogous metric, but I don't presently know enough about GBMs to be sure. | |
| Oct 4, 2013 at 20:42 | answer | added | Momo | timeline score: 46 | |
| Oct 4, 2013 at 19:32 | history | edited | Zach | CC BY-SA 3.0 |
added 650 characters in body
|
| Oct 4, 2013 at 19:26 | comment | added | Zach | GBM - generalized boosted model. I'm using the GBM package in R. | |
| Oct 3, 2013 at 22:54 | comment | added | Glen_b | Uh, and spell out what GBM stands for when you do, please? Is it Generalized Boosted Regression model? Something else? | |
| Oct 3, 2013 at 22:46 | comment | added | Glen_b | Could you put the additional things you're asking for in your actual question? | |
| Oct 3, 2013 at 17:42 | history | edited | Zach | CC BY-SA 3.0 |
added 373 characters in body
|
| Oct 3, 2013 at 17:33 | comment | added | Zach | Hmmm, so how would I calculate deviance for the GBM, in order to compare the models? Also, I'm trying to calculate this metric on a test set of new data, that wasn't used for fitting the model. | |
| Oct 3, 2013 at 15:08 | history | tweeted | twitter.com/#!/StackStats/status/385783409395924992 | ||
| Oct 3, 2013 at 3:02 | comment | added | Glen_b | Poisson regression at least should give you a deviance automatically | |
| Oct 3, 2013 at 2:59 | comment | added | Zach | A couple of different ways, ranging from a penalized poisson regression to a gbm. I'm looking for a good error metric to compare different models. Thanks for the advice. | |
| Oct 3, 2013 at 0:02 | comment | added | Glen_b | Deviance. How are you fitting your Poisson model? | |
| Oct 2, 2013 at 23:54 | comment | added | Zach | @Glen_b what's the formula for deviance? | |
| Oct 2, 2013 at 22:22 | comment | added | Glen_b | For Poisson models you could use the deviance; which is akin to a MSE but better suited to a Poisson. If you samples sizes aren't small, a weighted MSE would be quite similar. | |
| Oct 2, 2013 at 22:02 | history | edited | Zach | CC BY-SA 3.0 |
yup != yelp
|
| Oct 2, 2013 at 21:22 | history | edited | chl | CC BY-SA 3.0 |
edited title
|
| Oct 2, 2013 at 19:03 | history | edited | Zach | CC BY-SA 3.0 |
Added a note from kaggle
|
| Oct 2, 2013 at 18:56 | history | asked | Zach | CC BY-SA 3.0 |