- 最后登录
- 2016-11-27
- 注册时间
- 2003-1-21
- 威望
- 250
- 金钱
- 16832
- 贡献
- 11934
- 阅读权限
- 255
- 积分
- 29016
- 日志
- 4
- 记录
- 0
- 帖子
- 1438
- 主题
- 69
- 精华
- 0
- 好友
- 380
签到天数: 3 天 [LV.2]偶尔看看I - 注册时间
- 2003-1-21
- 最后登录
- 2016-11-27
- 积分
- 29016
- 精华
- 0
- 主题
- 69
- 帖子
- 1438
|
Tsai,除非有独特的理由,我看不见为什么要把大于1的Rwg算作0。这样做我觉得很奇怪。
关于Rwg和ICC的大小,我在南京已经给了一般的参考了。
对于Rwg: George, J. M., & Bettenhausen, K. (1990). Understanding prosaic behavior, sales performance, and turnover: A group-level analysis in a service context. Journal of Applied Psychology, 75, 698-709. (1)the Median RWG(J) across groups should be high. George and Bettenhausen (1990) proposed that RWG(J) greater than or equal to .70 could be considered as indicators of good agreement within group. (2)The percentage of groups with RWG(J) higher than .70 should be high.
关于ICC: LeBreton, J.M. & Senter, J.L. (2008). Answers to 20 questions about interrater reliability and interrater agreement. Organizational Research Methods, 11(4), 838-839. 他们的建议是: “We encourage researchers to adopt traditional conventions used when interpreting effect sizes (i.e., percentage of variance explained). Specifically, a value of .01 might be considered a ‘small’ effect, a value of .10 might be considered a ‘medium’ effect, and a value of .25 might be considered a ‘large’ effect (这是ICC1)…… So in most instances, ICC(K) and ICC(A,K) values >.90 are likely unnecessary. However, depending on the quality of the measures being used in multilevel analysis, researchers will probably want to choose between .70 and .85 to justify aggregation.”(这是ICC2) |
|