組內相關係數Z轉換信賴區間估計方法之研究

Abstract

在社會科學研究中，隨機效果模型是一個被普遍使用的模型，其中ICC (intraclass correlation coefficient) 為隨機效果模型的組內相關係數。若是以確切 (exact) 的方法在給定的信心水準(1-α) 、組內相關係數ICC(ρ) 以及組數 (observation) 下去計算所需的樣本數 (subject) ，實在是過於複雜。因此許多學者提出以大樣本Z轉換的近似法去計估計信賴區間的表現，也就是以常態分配的信賴區間法去近似。本篇論文是欲比較大樣本的近似方法好壞，並歸納在什麼時候適用這些近似方法。

In social science researches, the random effects model is commonly used and the ICC is the interclass correlation coefficient in this random effects model. It’s very complicated to use the exact method to calculate the required number of samples (subject) by given confidence interval (α), the correlation coefficient ICC (p) of the groups and the number of groups (observation). For this reason, some scholars put forward a method which use the confidence interval assigned by Z distribution in "simplified" method to estimate the required number of samples, that is to say they use the confidence interval assigned by normal distribution to approximate. This research will compare these different large sample size simplified methods and find out the proper using time of these methods.