Skip to content
Prev 12824 / 20628 Next

standard error and statistical significance in lmer versus lm

Dear prof. Bolker,

thanks a lot for your answer. I have not been able to observe any 
definitive pattern by following your suggestion:

One thing that might be interesting would be plotting the residuals from
a model of only probeset variation (i.e., values ~ (week|probeset))
and seeing how the week*genotype pattern was (hopefully) clarified.

Could you please be a little bit more specific about how the residuals 
from the model you suggest should be plotted?

Here the summary of the model, for yours and other readers perusal:
Linear mixed model fit by maximum likelihood  ['lmerMod']
Formula: values ~ (week | probesets)
    Data: dataset

      AIC      BIC   logLik deviance df.resid
172535.9 172590.6 -86263.0 172525.9   414340

Scaled residuals:
      Min       1Q   Median       3Q      Max
-11.2119  -0.5763  -0.0112   0.5812  21.3413

Random effects:
  Groups    Name        Variance Std.Dev. Corr
  probesets (Intercept) 5.166933 2.27309
            week        0.004412 0.06642  -0.25
  Residual              0.061336 0.24766
Number of obs: 414345, groups:  probesets, 18015

Fixed effects:
             Estimate Std. Error t value
(Intercept)  6.64697    0.01662     400


Thanks again,

Vincenzo
On 12/27/2014 4:33 AM, Ben Bolker wrote: