Timeline for LMM Results interpretation: Change in results when adding interaction
Current License: CC BY-SA 4.0
5 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Aug 1, 2022 at 9:35 | comment | added | Juliette | this is where I get confused, because it's when I look at the significance of a predictor with the Anova function, it is there that I see A different X2 and p value for proximity depending on my reference level of dyad_type. This is what gets me confused, not the change in the individual levels of my covariate but the change in the predictor in itself, and I'm not sure which results to rely on. I hope this makes sense? | |
| Jul 31, 2022 at 13:15 | vote | accept | Juliette | ||
| Jul 31, 2022 at 13:06 | comment | added | EdM |
@Juliette it doesn't matter which reference level you choose. The overall model is the same regardless, as are any predictions you make from the model. The "significance" of an individual coefficient for a predictor involved in interactions doesn't matter. To evaluate the significance of a predictor in an interaction, you need to do a combined test of its individual coefficient along with the coefficients of its interaction terms. That's done for example by the Anova() function of the R car package.
|
|
| Jul 31, 2022 at 7:57 | comment | added | Juliette | thank you for your answer! I'm not completely sure I understand, even if it's normal that the reference level changes my coefficients and significance levels, in this case, which final result should I use? (how do I chose which reference level to stick to? ) | |
| Jul 30, 2022 at 17:04 | history | answered | EdM | CC BY-SA 4.0 |