Timeline for Is it possible (or even advisable) to only include random slopes for some contrasts of levels of a factor but not others?
Current License: CC BY-SA 4.0
4 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Nov 20, 2018 at 18:40 | comment | added | Dave | Hi Isabella, here is a link to the paper! arxiv.org/abs/1506.04967 Thanks once again. | |
| Nov 20, 2018 at 2:16 | comment | added | Isabella Ghement | Thank you very much for your update, @Dave! I am not familiar with the use of PCA on the random effects structure - can you give me a reference for it? As long as you can clearly justify your choice of which contrast is best supported by the data, you should be fine. | |
| Nov 20, 2018 at 0:51 | comment | added | Dave | Hi Isabella, thank you for the reply. I'd determined that (for instance) m2 is preferable by looking at the amount of variance associated with each slope when they are both fit. After using PCA on the random effects structure, it was determined that only two random subject effects should be fit. These two factors together suggest that a slope for only one contrast should be fit, and which it should be (based on variance associated with each). | |
| Nov 10, 2018 at 4:22 | history | answered | Isabella Ghement | CC BY-SA 4.0 |