$ c! j) }6 c8 s$ u. ~) rkenny,我的问题就如如下这段话所述(该段话来自Liu et al., 2012, AMJ): To justify that HLM3 was appropriate for analyzing our three-level data, we first ran null models with no predictors but creativity as the dependent variable (Raudenbush et al., 2004). The results show significant between-team variance ( x2 (86) =215.68, p< .001; ICC1 =.28 [indicating 28 percent of variance resides between teams]) and between-department variance ( x2(21)=112.43, p <.001; ICC1 = .18 [indicating 18 percent of variance resides between departments]) in creativity. : g$ Z- s/ K8 L- }6 K
该文算出来的between-department variance of creativity 是显著的,ICC1(department)=0.18.但是我的data算出来between-department variance of creativity不显著,相应地ICC1(department)也比较小。reviewer质疑我的between-department variance of creativity都不显著,我为何还要在department level 放preditor呢??但是我放的departmen-level predictor对creativity的回顾系数是显著的,我该如何respond?? 希望我这次讲清楚了我的问题。作者: Kenneth 时间: 2013-3-27 12:26
组间方差不显著,代表没有什么可以被估计,那就没有必要找 predictor了,reviewer是对的。. Y) R, F7 K/ M' p: Q1 D: k+ R U
比如,如果当 x=0 时,y一定等于.01; 当 x=1时, y 一定等于.02。你问我y=.01 和 y=.02 有没有分别,我说微乎其微。但是x与y的相关是多少呢? rxy=1.0*** !