Timeline for u-shape for logistic regression?
Current License: CC BY-SA 4.0
12 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Dec 11, 2018 at 18:20 | comment | added | AdamO | @GreenPirate As I said, the LOESS smooth suffices to explore agnostic functional forms of the mean response. | |
| Dec 11, 2018 at 18:18 | comment | added | GreenPirate | @AdamO "You're done" always sounds good. ;) I see e.g. in this article a pledge to do more (onlinelibrary.wiley.com/doi/abs/10.1002/smj.2399) and also examine the ends and finding the maximum. Do you mean by "other models" the GAM model or the other quadratic term variations? | |
| Dec 11, 2018 at 15:09 | comment | added | AdamO | @GreenPirate If you want to test for the presence of a U-shape trend, you already found it: the statistical significance of the linear and quadratic term is there so you are done. These other models you fit are useless because they do not relate to any question you have asked beforehand. | |
| Dec 11, 2018 at 8:50 | comment | added | GreenPirate | @AdamO what would you do then? I also tried to use a log of wine and it turns out to be significant. However, I would like to test on shapes more systematically than right now. | |
| Dec 6, 2018 at 20:53 | history | edited | AdamO | CC BY-SA 4.0 |
added 408 characters in body
|
| Dec 6, 2018 at 17:15 | comment | added | AdamO | @GreenPirate think about the model. You know a straight line doesn't fit the trend well. Neither does a quadratic whose apex is constrained to be at 0. Always plot the predicted trend. Practice safe stats. | |
| Dec 6, 2018 at 9:01 | comment | added | GreenPirate | Thank you both. I added a wine squared (so "glm(dv_death ~ wine + winesq + cigarettes + (...)) and both wine and winesq are significant in this model. However, when I use a limited model (either dv_death ~ wine OR dv_death ~ winesq), none of them is significant. | |
| Dec 5, 2018 at 18:57 | comment | added | ColorStatistics | I can't be confident that the inclusion of the quadratic term is justified based on the plot of death vs. wine. In my mind, that is not the immediately relevant plot to analyze in order to decide whether or not the inclusion of a quadratic term or splining is justified. On the other hand, if we saw a quadratic relationship between the log of odds and wine, then I'd be confident that we're justified to include a quadratic term. | |
| Dec 5, 2018 at 18:45 | comment | added | AdamO | @ColorStatistics I think I see your point. I am borrowing on my prior knowledge in this area. You are right that, in general, for GLMs a non-linear trend in the expected response doesn't necessarily mean there's a non-linear trend in the linear predictor. HOWEVER, the inverse-logit is monotonic, and yet the trend reverses direction, so an inflection point is present. | |
| Dec 5, 2018 at 18:32 | comment | added | ColorStatistics | If this were a linear regression then the observed u shape between wine and death may justify inclusion of a quadratic term. However, given that this is a logistic regression and the dependent variable is the log of the odd of death, why would a quadratic relationship between wine and death justify the exploration of a quadratic relationship between the log of odds of death and wine? In my mind, there is a disconnect here. | |
| Dec 5, 2018 at 18:01 | history | edited | AdamO | CC BY-SA 4.0 |
added 164 characters in body
|
| Dec 5, 2018 at 17:30 | history | answered | AdamO | CC BY-SA 4.0 |