Eiko Fried
2012-Apr-12 12:41 UTC
[R] Multivariate multilevel mixed effects model: interaction
Hello. I am running a multivariate multilevel mixed effects model, and am trying to understand what the interaction term tells me. A very simplified version of the model looks like this: model <- lmer (phq ~ -1 + as.factor(index_phq) * Neuro + ( -1 + as.factor(index_phq)|UserID), data=data) The phq variable is a categorical depression score on 9 depression items (classified by the variable "index_phq"), and Neuroticism is a covariate (random effect = the subjects of my study, they should be allowed to vary in slope/intercept). The output looks like this: Fixed effects: Estimate Std.Error t-value as.factor(index_phq)1 0.0185335 0.0466784 0.397 as.factor(index_phq)2 -0.0327630 0.0453493 -0.722 as.factor(index_phq)3 0.2210580 0.0612363 3.610 as.factor(index_phq)4 0.6240455 0.0582654 10.710 as.factor(index_phq)5 0.1121727 0.0622298 1.803 as.factor(index_phq)6 -0.1797495 0.0484940 -3.707 as.factor(index_phq)7 -0.0907630 0.0488517 -1.858 as.factor(index_phq)8 -0.0854370 0.0384632 -2.221 as.factor(index_phq)9 -0.0621792 0.0319729 -1.945 Neuro 0.0218704 0.0019690 11.107 as.factor(index_phq)2:Neuro 0.0026006 0.0018268 1.424 as.factor(index_phq)3:Neuro 0.0012185 0.0025774 0.473 as.factor(index_phq)4:Neuro 0.0009458 0.0023336 0.405 as.factor(index_phq)5:Neuro 0.0061280 0.0025547 2.399 as.factor(index_phq)6:Neuro 0.0070453 0.0018788 3.750 as.factor(index_phq)7:Neuro -0.0008890 0.0020755 -0.428 as.factor(index_phq)8:Neuro -0.0092411 0.0020197 -4.575 as.factor(index_phq)9:Neuro -0.0159498 0.0021470 -7.429 Looking only at the interaction effects, do I understand correctly that ... 1) Neuroticism has a significant effect on all PHQ items, because the effect is significant on phq1 (t-value=11.107), and the other t-values are at a maximum -7.429 lower than the t-value of 11.107, meaning they are still significant because they are still all above t~2? 2) The strongest effect is on PHQ6, which is significantly higher than the effect from Neuro on PHQ1? 3) The weakest effect is on PHQ9, which is significantly lower than the effect from Neuro on PHQ1? Thank you Eiko [[alternative HTML version deleted]]
Possibly Parallel Threads
- Oddity: I seem to have a variable in a dataframe that doesn't show in colnames() - can anyone advise?
- anova of lme objects (model1, model2) gives different results depending on order of models
- reshape and geeglm problem
- vorbis physical bitstream structure
- Comparing dcast and reshape