[R-sig-ME] Do I need random effects?

[Murray Jorgensen]

I want to test if two random effects are needed. I rather hope that they 
are not.

I believe the 0/1 response mincr [male increase] should depend on the 
factor icfac [initial configuration] but not on two other factors gap 
and comp, which would be crossed were it not for missing values.


 > m1 = lmer(mincr ~ icfac + (1|gap) + (1|comp), REML = FALSE)
 > summary(m1)
Linear mixed model fit by maximum likelihood
Formula: mincr ~ icfac + (1 | gap) + (1 | comp)
     AIC    BIC logLik deviance REMLdev
  -259.8 -228.7  136.9   -273.8  -250.3
Random effects:
  Groups   Name        Variance Std.Dev.
  comp     (Intercept) 0.000000 0.00000
  gap      (Intercept) 0.000000 0.00000
  Residual             0.037834 0.19451
Number of obs: 627, groups: comp, 40; gap, 8

Fixed effects:
               Estimate Std. Error t value
(Intercept)  1.261e-11  1.108e-02   0.000
icfacmale1   5.556e-01  2.331e-02  23.836
icfacfem1   -3.129e-11  1.715e-02   0.000
icfacother   2.500e-01  6.966e-02   3.589

Correlation of Fixed Effects:
            (Intr) icfcm1 icfcf1
icfacmale1 -0.476
icfacfem1  -0.646  0.307
icfacother -0.159  0.076  0.103

Zero variance components are estimated for the gap and comp effects. 
Goodie! But does this constitute a test? What if they were merely smallish?

Then I would like to do:

m0 = lmer(mincr ~ icfac)
anova(m0,m1)

But lmer won't let me fit models with no random effects.

Regards,  Murray

-- 
Dr Murray Jorgensen      http://www.stats.waikato.ac.nz/Staff/maj.html
Department of Statistics, University of Waikato, Hamilton, New Zealand
Email: maj at waikato.ac.nz    majorgensen at ihug.co.nz      Fax 7 838 4155
Phone  +64 7 838 4773 wk    Home +64 7 825 0441   Mobile 021 0200 8350