Object To declare the distribution of intra-unit correlation coeffient (Ï) of repeated measure data . Methods Monte Carlo method is used to get random sample from the populations with assumed parameters. Restricted Iterative Generalized Least Squares(RIGLS) is used to estimate the parameters of the level-2 model, the intra-unit correlation coeff ient (Ï) of every bootstrap sample is estimated based on the model. Results The type of sampling distribution varies with the Ï.When Ï is less than 0.5, Ï is distributed which has positive skewness, When larger than 0.5, Ï is distributed which has negative skewness , and about 0. 5, then distribution is normal. The standard error of Ï is proportion to the sample size.With the same total sample size, the larger the number of level-2 unit is ,the smaller the SE of intra-unit correlation coeffient will be. Conclusion The distribution of intra-unit correlation coeffient of repeated measure data always mirrors the dependence of the data. The results of interested can act as a guide in the further analysis of repeated measure data.
|