[R-sig-ME] Questions of nlme package in R

[Ben Bolker]

On Sun, Dec 11, 2016 at 9:14 PM, 欧阳湘子 <oyangxz0412 at 163.com> wrote:
> Dear professor，
> hello，my name is Sharon. I have a question about the nlme package in R. When apply this package to do multilevel model，the function is “lme”. Why the summary part didn't show the significance test of the random effect？
> I will appreciate that you could answer my questions. Looking forward to your reply，thank you！
>
>
> regards，
> Sharon
>
>

  (Sorry; I replied privately that this should go to the list but
didn't see that you had *already* sent it to the list.)

   If you read Pinheiro and Bates 2000 (Mixed-effect models in S and
S-PLUS) you'll see that in the modern mixed model framework, testing
variance components is a little tricky; typically you can use
likelihood ratio tests, but they are known to be conservative.
http://tinyurl.com/glmmFAQ#testing-significance-of-random-effects also
has a bit of information, although it doesn't give the basic recipe:
to test the significance of a random effect, fit the model with (call
the fitted model m1) and without (m0) the random effect, and use

anova(m0,m1)

to perform a likelihood ratio test.

  sincerely
    Ben Bolker