組內相關係數之假設檢定最適樣本數

Sample size Calculation for Hypothesis Testing of Intraclass Correlation Coefficients (2)

楊乃蓁(Nai-Chen Yang)

指導教授 ： 謝國文

國立交通大學/管理學院/管理科學系所/碩士(2016年)

摘要

隨機效果模型中的組內相關係數ICC (intraclass correlation coefficients)在實際應用上無所不在，ICC(2)的最適樣本數如何決定，為本文章主要貢獻。一般較常見的指數有皮爾森相關係數(Pearson correlation coefficient)，雖然藉此可獲知變數之間的相關性，但皮爾森相關係數需滿足的前提假設較為困難，必須假設兩變數為相互獨立，所以在實際應用上較為困難，實際研究中資料多為有相似性且不獨立數據，於是使用組內相關係數來測量其變數之間的關係較為恰當。本文將專注研究於單因子隨機效果，使用統計軟體SAS/IML程式進行模擬，透過假設檢定法在檢定力及顯著水準α固定下，運用F分配作檢定找到最適樣本數。

關鍵字

組內相關係數 ； 假設檢定 ； 最適樣本數

並列摘要

In the practical applications, Intraclass correlation coefficients is the model in random effects model. The main contribution in this research is how to determine the optimum number of samples based on ICC(2). For the begging we learn to use the tools of measurement is the Pearson correlation coefficient. Although, we can know the correlation between variables, but satisfy the assumption is difficult in the reality. The reason is that the two variables of Pearson correlation coefficient t must be assumed independent. Therefore, in practical application is not easy. Most of the research data have similarities and dependent. Due to the reason above, Using the correlation coefficient ICC(2) to measure the relationship between variables is more appropriate. This article will focus on research on One-way random effects model, using statistical software SAS / IML program do the simulation. Through the hypothesis testing method to set the power, and employ the F ratio to look for the optimistic sample size.

並列關鍵字

intraclass correlation coefficients ； hypothesis tests ； optimal sample size

參考文獻

李仁豪, & 余民寧. (2008). 二層次結構方程式模型的應用: 以教育心理學為例. 師大學報, 53(3), 95.

Bartko, J. J. On various intraclass correlation reliability coefficients. Psychological Bulletin, 1976, 83, 762- 765.

Bonett, D. G. (2002). Sample size requirements for estimating intraclass correlations with desired precision. Statistics in medicine, 21(9), 1331-1335.

Donner, A., & Eliasziw, M. (1987). Sample size requirements for reliability studies.Statistics in medicine, 6(4), 441-448.

Gardner, M. J., & Altman, D. G. (1986). Confidence intervals rather than P values: estimation rather than hypothesis testing. BMJ, 292(6522), 746-750.