- 最后登录
- 2015-6-23
- 注册时间
- 2009-3-30
- 威望
- 0
- 金钱
- 139
- 贡献
- 82
- 阅读权限
- 20
- 积分
- 221
- 日志
- 0
- 记录
- 0
- 帖子
- 28
- 主题
- 9
- 精华
- 0
- 好友
- 0
  
- 注册时间
- 2009-3-30
- 最后登录
- 2015-6-23
- 积分
- 221
- 精华
- 0
- 主题
- 9
- 帖子
- 28
|
Kenney,
! o5 H3 a2 Q7 W0 h2 v
& V3 U" W4 G, n: M 您好!
6 A' q, i* G! K' n 首先非常感谢您在中人网上给我非常详细的解答,我感觉受益匪浅!
9 B, Q# ]% x0 f: a) q# p; N 但是同时也让我陷入了矛盾之中,如果按照我的模型修正方法,把同一构念的两个观测指标的测量误差相关,整体模型的拟合度将会显著提高,而且研究的关键路径 也会显著!如果不这样做,可能这篇paper就要废掉了,我有点矛盾!我见过别人这样调整模型,但是在正文中没有报告出来,模型 的拟合度很高,也发在了级别很高的期刊上!
7 `3 J* N* X$ o3 V7 d& b 谢谢,祝您一切顺利!
: v7 x" [5 A* E$ Y& J+ ` Tonylev 9 i' V. a! J6 a P& s4 l
- B7 E% w' Y0 `6 o" L; w& c/ ] r0 m l, C0 i0 ^& d
9 f/ R/ A+ ^$ I; z) ]5 B9 o0 r
* q4 o3 ]5 @$ g% E
Hi Tony,
7 g% u/ d2 \- z* i" E% W+ x7 ^0 M9 c" l# L7 o4 y j
What don't you reply on the blog? I would hope that we can carry on with the discussion there so that everyone can learn together.
$ B) Y/ T; [+ o$ r. e0 ~1. How do you know they do so if they do not report?
2 s4 }* ~& @. c8 I! d! j2. If there is nothing wrong with it, why don't you report?$ k, W, ~9 i3 |# }: M; @# L9 G% |" f
3. If you know that you cannot report (otherwise you would be challenged), therefore, you keep silent about it. Is this academic dishonest?2 ~" s: d" f; N8 }
4. If your fit would increase significantly simply be correlating ONE measurement item, that means it is the measurement model which causes the problem, not your structural model. I would suggest two approaches to you:4 e& }0 _' N3 a2 H# _) U$ X# o
(a) use item parceling to group the two correlated items; 5 z9 B5 V6 N$ D. \# G
(b) first run a CFA, then average the items for each construct and run a path analysis (no measurement model)./ O2 D& j* D( F: V5 O; w
9 B- n. `6 q! I. {Kenny! W8 M9 {* S& C) E# B& H
|
|