1 Introduction

The data set klein2000{multilevel} contains 750 observations from 50 groups.

2 Data Management

data(klein2000, package = "multilevel")
dta <- klein2000[, -c(3, 4, 7, 8, 9)]
head(dta)
  GRPID    JOBSAT       PAY    NEGLEAD
1     1 -1.717411  0.510543  0.3127974
2     1  0.237316 -1.967199 -1.9046836
3     1 -2.097958 -2.748838 -0.1243344
4     1  2.932552  1.355283 -1.2983974
5     1  0.144981 -1.185332  0.0646287
6     1 -1.995481 -0.399117  0.3870573

Column 1: GRPID ~ numeric Group Identifier
Column 2: JOBSAT ~ numeric Job Satisfaction (DV)
Column 3: PAY - numeric Pay
Column 4: NEGLEAD - numeric Negative Leadership

# 計算變數
dta %>%
  mutate(NEGLEAD_mt = mean(NEGLEAD))%>%
  group_by(GRPID) %>% 
  mutate(mneglead=mean(NEGLEAD), NEGLEAD_mgc=mean(NEGLEAD) - NEGLEAD_mt,
         mpay=mean(PAY), PAY_c=PAY - mean(PAY)) -> dta

3 Models

3.1 Model 1

m0 <- lme4::lmer(JOBSAT ~ (1 | GRPID), data=dta)
m1 <- lme4::lmer(JOBSAT ~ PAY_c + (1 | GRPID), data=dta)
sjPlot::tab_model(m0, m1, show.p=FALSE, show.r2=FALSE, show.obs=FALSE, show.ngroups=FALSE, show.se=TRUE, show.ci=FALSE)
  JOBSAT JOBSAT
Predictors Estimates std. Error Estimates std. Error
(Intercept) 0.08 0.14 0.08 0.14
PAY c 0.97 0.08
Random Effects
σ2 5.47 4.51
τ00 0.68 GRPID 0.75 GRPID
ICC 0.11 0.14

結果顯示：

  • 有4.71% [((5.47+0.68)-(4.51+0.75))/(5.47+0.68)]的工作滿意度變異可由同組內個人薪資差異加以解釋。

3.2 Model 2

m2 <- lme4::lmer(JOBSAT ~ NEGLEAD_mgc + (1 | GRPID), data=dta)
sjPlot::tab_model(m0, m2, show.p=FALSE, show.r2=FALSE, show.obs=FALSE, show.ngroups=FALSE, show.se=TRUE, show.ci=FALSE)
  JOBSAT JOBSAT
Predictors Estimates std. Error Estimates std. Error
(Intercept) 0.08 0.14 0.08 0.12
NEGLEAD mgc -1.78 0.34
Random Effects
σ2 5.47 5.47
τ00 0.68 GRPID 0.31 GRPID
ICC 0.11 0.05

結果顯示：

  • 有54.41% [(0.68-0.31)/0.68]的組間的工作滿意度變異可由組間的負向領導力差異加以解釋。