評估取樣權重於檢定Likert問卷之測量恆等性研究

Abstract

本研究的目的主要探討應用取樣權重（sampling weights）於檢定Likert問卷間之測量恆等性（Measurement Equivalence, ME）時，對於檢定正確性的影響。在大型調查研究中常因不同群體中個體被抽中機率不相等，而必須適當加上取樣權重，才能正確的反應出母群體資料。因此，在進行大型調查研究時，取樣權重對於測量恆等性檢定的影響就顯得格外重要，但也容易被研究者所忽略。本研究利用MIMIC模式進行Likert問卷之測量恆等性檢定時，取樣權重對於檢定正確性影響。研究中，採用五點計分Likert問卷模式，實驗設計包含了多種不同的取樣權重及抽樣的樣本數。研究結果發現進行測量恆等性檢定時，在不同研究設計下，使用取樣權重相對於未考慮取樣權重能獲得較佳的檢定正確性。

This study is aimed to examine the weighting effects introduced by stratified sampling procedures on the accuracy of measurement equivalence (ME) between Likert- type questionnaires. To achieve the population- wise generalizability successfully, proper weights have to be engaged in all levels of statistical analyses. In identifying measurement inequivalence between Likert- type questionnaires of large scale surveys, the influences of weights are as critical as any other studies yet they are often neglected. A numerical simulation study is conducted to examine the accuracy of analyzing ME when using MIMIC model approaches under experimental designs of various weights and population sizes. The results demonstrate the weighting effects on accuracy of ME identification are considerable and practical guidelines are provided for research communities.