Clinical Science

Publisher Website
Google Scholar
Abstract
Analyses of variance under the assumptions of either independence or intraclass correlation among observations were carried out on theoretical dental research data. Data sets varied with respect to the magnitude of a between-subjects treatment effect and the variance of scores within subjects. When scores within subjects were positively correlated (i.e., the intraclass correlation coefficient was greater than zero), the independence model provided tests which were too likely to reject the null hypothesis. Dental researchers who analyze sites (e.g., teeth, pockets) as independent observations, without establishing the intraclass correlation to be near zero, run the risk of reporting results in terms of an improper statistical model

This publication has 6 references indexed in Scilit: