0
$\begingroup$

I have two nested mixed models: the difference in these models is the presence of one predictor variable (type). I used lmerTest to obtain p-values for the model, and none except the intercept were significant. Can I still compare these models using anova to say the predictor variable type is (or is not) having an effect on the model?

m1 <- lmer(f2 ~ sex + type + (1|speaker))
m2 <- lmer(f2 ~ sex + (1|speaker))
anova(m1, m2)
Improve this question
$\endgroup$
6
  • $\begingroup$ Have you already run this code? Do you get a significant or a non-significant difference? $\endgroup$
    – amoeba
    Commented Oct 19, 2016 at 22:45
  • $\begingroup$ I have run this code (my actual code has a few more predictor variables, though) and get a p-value of 0.9885, which I take to mean that these models are not significantly different and therefore it is unlikely that the predictor variable type has a significant effect $\endgroup$
    – Lisa
    Commented Oct 19, 2016 at 22:49
  • $\begingroup$ That's right, this means that the difference is not statistically significant (whether it is "likely" that type has an effect or not, we cannot really say; it depends on how "likely" you thought it was before running your experiment and on many other things). So in your case the conclusions of lmerTest and of anova agree. I am not sure what exactly is then your question. $\endgroup$
    – amoeba
    Commented Oct 19, 2016 at 22:49
  • $\begingroup$ My question is: Can I compare these mixed models despite the fact that the individual models do not show any significant effects? $\endgroup$
    – Lisa
    Commented Oct 19, 2016 at 22:52
  • 1
    $\begingroup$ Sure. This question suggests that perhaps you do not quite understand what is going on here. Comparing m1 and m2 is equivalent to testing whether type is a significant predictor in m1. So running anova(m1, m2) and running lmerTest on m1 are just two different ways to test if type is significant. You can use either of these two ways, or both, if you like. You are testing the same thing two times. Does it make sense? $\endgroup$
    – amoeba
    Commented Oct 19, 2016 at 22:58

0

Reset to default

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.