nlme和lme4中的不同随机效应

Question

I fit this model in nlme :

library(nlme)
data("Machines")
fit1 <- lme(score ~  - 1 + Machine, random=~1|Worker, data=Machines)

I can get the coefficients with

> fit1$coefficients
$fixed
MachineA MachineB MachineC 
52.35556 60.32222 66.27222 

$random
$random$Worker
  (Intercept)
6 -8.70711058
2 -1.59425968
4 -0.06931564
1  1.21035769
3  6.21174760
5  2.94858062

now I fit the same model in lme4

library(lme4)
fit2 <- lmer(score ~  -1 + Machine + (1|Worker), data=Machines)

I get the exact same fixed effects:

 >summary(fit2)
 ...
 Fixed effects:
          Estimate Std. Error t value
 MachineA   52.356      2.229   23.48
 MachineB   60.322      2.229   27.06
 MachineC   66.272      2.229   29.73
 ...

I now want the random effects per worker, they aren't displayed in the summary, but it must be this:

 > fit2@u
 [1] -5.34898187 -0.97939105 -0.04258222  0.74355106  3.81602197    1.81138210

why are they different from the nlme results while the fixed effects are the same?

Answer 1

Use ranef() to extract the random effects.

library(lme4)
library(nlme)
data("Machines")

fit1 <- lme(score ~  - 1 + Machine, random=~1|Worker, data=Machines)
ranef(fit1)

fit2 <- lmer(score ~  -1 + Machine + (1|Worker), data=Machines)
ranef(fit2)