- 最后登录
- 2014-9-15
- 注册时间
- 2013-11-14
- 威望
- 0
- 金钱
- 65
- 贡献
- 40
- 阅读权限
- 20
- 积分
- 105
- 日志
- 0
- 记录
- 0
- 帖子
- 17
- 主题
- 4
- 精华
- 0
- 好友
- 0
- 注册时间
- 2013-11-14
- 最后登录
- 2014-9-15
- 积分
- 105
- 精华
- 0
- 主题
- 4
- 帖子
- 17
|
罗老师好,想请教您一个问题: y; D9 n( y, \- G* N
' N6 K5 O1 J( b) e+ B" T3 h我在Leadership Quaterly上面看到一篇文章,是用MSEM测试的 一个2-1-1-2 sequential mediation. / E3 x4 k7 T4 [" G5 J3 \% ~7 p
( t% P5 q+ ?0 Z* g
Nohe, C., Michaelis, B., Menges, J. I., Zhang, Z., & Sonntag, K. (2013). Charisma and organizational change: A multilevel study of perceived charisma, commitment to change, and team performance. The Leadership Quarterly, 24, 378-389.
. t! n5 ?4 @8 Y0 t1 y
- J: n3 ?0 I& I4 u' b- _ o! ]- O在处理cross-level mediation的时侯,作者
! h- P$ ~3 I# y“decompose the variance of a variable into its latent within-unit component (within-team variance) and a latent between-unit component (between-team variance; Lüdtke et al., 2008; Muthén & Asparouhov, 2009). Relationships between these variance components can be modeled independently at each level through the specification of structural and measurement models.”[p.383]( B7 ~7 @/ L; E
- L; x b1 U% o% W0 ^
按照这个方法,我用Mplus把level 1的variable 分别在Level 1和Level 2作为latent variable 建Model (两个level 1的variable,每个都由7个item来衡量),然后只用Level 2的coefficients来计算the indirect effect of sequential mediation. ( \2 D6 V) [, A. S+ T7 y6 }5 ?
2 I1 @0 D* J- Q7 E" E W' Z/ [
虽然Mplus给出了coefficient的estimation和significance level,但是model整体“nonidentified”,然后Mplus report “THE NONIDENTIFICATION IS MOST LIKELY DUE TO HAVING MORE PARAMETERS THAN THE NUMBER OF CLUSTERS. REDUCE THE NUMBER OF PARAMETERS.”
7 j5 U- j6 _5 r; U: r# _+ Q9 r6 `5 p# {
我想这大概是主要因为每个latent variable的measure item比较多,可是group number只有39个的原因(虽然individual level sample 有1800多个)。
1 Q9 d3 m ~! l: q5 D& _3 j" {0 X* h! a
我试了另外一个model,不是用latent variable,而是用直接的aggregated variable分别在Level 1(mean of the 7 measure items) 和Level 2 (group average of the level 1 variable)参与模型运算。还是只看所有Level 2的coefficients来计算2-1-1-2的indirect sequential mediation。 Mplus就没有任何warning了。
& R- F! {6 V' `$ z. ~1 ?# i
; Y3 y7 }. d( K4 O9 c我想请问罗老师,如果我在limitation里面承认model有不完美的地方,我可以用我自己试的方法,不用latent variable,而是直接在Level 1和Level 2用aggregated value来计算一个2-1-1-2的模型么? ' g- x- |' c! s( c5 q4 A9 |
9 N5 K8 \/ g ^' }5 n
真诚希望得到您的指点!
, X: `6 G* z, j3 e7 H
( S/ V9 ?, v9 G谢谢先!7 z0 v }+ b( Z% W E% A9 E
@' ^* }! J) e& u
. L( V9 P, c2 v5 h# `7 Q$ k% X
8 @$ l) b+ \! d- Y' A. C7 {/ i
/ X. }3 m4 U; e
& W u8 T4 S. Z/ K) j+ d, H& D. l, d7 v8 I+ S
|
|