# Thread: Mixed effects using nlme in R

1. ## Mixed effects using nlme in R

If I include an interaction (time*score) in my model the main effect (score) is significant. However, if I remove the covariate (time), the effect for (score) is no longer significant. Can anyone give a simple explanation why?
Thanks

2. ## Re: Mixed effects using nlme in R

hi,
maybe others will reflect as well, but my simple explanation would be that as follows: the significance is decided based on the residual variance, effect size and sample size, roughly. If time is a covariante then a part of the variation is attributed to the time variable, so the residual variation is less and you might be able to see the effect of the score. If you take out the time variable then the variance explained by time is now increasing the residual variation - and the increased residual variation might be too large to make the effect of score visible.

regards

3. ## Re: Mixed effects using nlme in R

Please write out your models for your above scenarios. If the interaction is significant, it is telling you that score effect on Y is conditional on time:

y-hat = intercept + time + score + time*score

4. ## Re: Mixed effects using nlme in R

but why does the choice of covariate influence the effect of the other variables outside of any interaction effect. I have two models

lmer(X ~ A*B*C, (1|R), data)
lmer(X ~ A*B*Z, (1|R), data)

The A*B interaction is significant when C is is included but not when Z is included (and the estimate of effect is considerably more). Please note, I'm not referring to the A*B*C or A*B*Z interactions.

 Tweet

#### Posting Permissions

• You may not post new threads
• You may not post replies
• You may not post attachments
• You may not edit your posts