組內相關係數區間估計之樣本數方法

Abstract

不論是在自然或是社會科學中，衡量組間變異與組內變異的關係程度，常會提及組內相關係數的概念，其能夠反映出資料的組間差異性，這個指標常用在檢驗變數的變異能夠透過群組變異解釋的程度，可評估個體資料是否適合彙總作為群組資料，或是評估對同一受測者重複測量的結果是否具有一致性，然而在相關研究過程中，樣本數決定是重要的一環，因此本文以SAS/IML軟體建立模型進行模擬，探討在設定信心水準、施測次數、組內相關係數條件下，以區間寬度法與區間涵蓋率法作為兩大模擬情境，計算出能夠滿足條件情境的最小要求樣本，並且對求得樣本數大小進行分析比較。

Whether in the natural sciences or the social sciences, intraclass correlation coefficient is always used to measure the degree of the correlation between the interclass variation and the intraclass variation. This index reflects the differences between groups of data. It’s always regard as a way to inspect variation of variable is interpreted by variable between groups for evaluating whether integrating individual data to group information. Also, intraclass correlation coefficient is used to inspect the consistency of retest measured results. In this dissertation, the SAS/IML software is used to construct the model, and there are two simulation situations, interval width and interval coverage rate, to find the optimal sample size in conditions of conditional level, measurement and intraclass correlation coefficient. These calculated sample size are analyzed and compared.