The Effects of Nonresponse Rates on Group-Level Correlations
研究群体成员无应答如何影响群体层面变量间的相关性，发现组内相关系数(ICC)水平是关键调节因素，高ICC下低应答率也能得到准确估计，而低ICC下高应答率也可能导致相关性衰减。
This study investigates the effects of group member nonresponse in research that examines relationships between group-level constructs aggregated from the responses of individual group members. Simulation data show that the effects of nonresponse vary greatly depending on levels of between-group variability and within-group variability expressed jointly as intraclass correlation, or ICC(1), values. Even high response rates, for example, can produce significantly attenuated correlations between group-level variables with ICC(1) levels commonly reported in group studies. Conversely, even low response rates can yield accurate group-level correlations when ICC(1) levels are high. These effects of nonresponse are equivalent whether nonresponse derives from a random or biased process. Group size appears to play no important role in moderating the effects of response rate. Suggestions for research design and post hoc diagnosis of findings are discussed.