Hello,
I have an LMM with one fixed factor that has a significant effect. However, items in the two levels of the fixed factor differ significantly on a continuous variable. I entered that variable into the model and found that it had a significant effect too.
Is there a way to determine whether the unique contribution of the fixed factor is significant beyond the contribution of the continuous variable?
Thanks,
Chen
Hi Chen,
not sure if I understand correctly, but I think once you enter the possible confounding variable into your model, you are effectively controlling for it, i.e., the remaining fixed effect of your dichotomous factor is due to that factor only, with the continuous confounder being kept constant. So you can test the significance of the fixed effect like you would normally do (with all caveats which apply in a mixed-effects model).
Best regards, Lars
When both are in the model, they are competing for the same shared variance which increases the uncertainty in the parameter estimates. The fact that it's still significant is a good sign. However, some reviewers might want you to first run a model with the confounding variable, then a model with both variables, and then test if the second is better than the first (in terms of adjusted R2, AIC, BIC, etc.). That hierarchical regression approach is commonly used to assess "unique" variance accounted for.
- Badges
- Science topic
- Similar topics
- Correction