使用李克特式資料之組內相關係數分析

Abstract

社會科學研究中常使用問卷的方式獲取資料以分析研究，但李克特式資料相較於一般常態性資料會有資訊遺漏的情形發生，導致會影響到問卷分析時之顯著水準及檢定力，本文探討在隨機效果下，樣本數對組內相關係數檢定的影響，主要方向為需要多大的樣本數才會使檢定的效果及正確性和常態資料相當。

In social science research, we often get data from questionnaires to conduct a series of analysis. There are more or less missing data while we use the likert-type scale data, resulting the biases of the significant level and the power. This paper aims to discuss what the influence of sample size would bring to the test of intraclass correlation coefficient in the random effects model. The main direction is to figure out the simple size which can enable the effects of the test and accuracy to get pretty closer to the normality data.