無差異檢驗的兩種方法比較：Schuirmann無差異檢驗和student's t檢驗

Comparison between Two Approaches for Equivalence: Schuirmann's Test of Equivalence and Student's t

李中權(Zhong-Quan Li) ； 王力(Li Wang)

《αβγ量化研究學刊》 2卷1期 (2008/12) Pp. 126-138

摘要

在心理與教育研究中，有時研究者也會對兩種實驗處理產生的影響或者兩組實驗物件自身的特性是否存在實質性差異感興趣。採用傳統虛無假設檢驗（如Student's t 檢驗）來做判斷可能會導致一些錯誤，而借助無差異檢驗的方法（如Schuirmann無差異檢驗）進行處理會更合適。本研究通過操控樣本容量、樣本量比例、總體方差比例以及總體均數差異構成，模擬生成研究中常見的資料狀況，比較了Schuirmann無差異核對總和student's t檢驗在探測兩組均數是否存在實質性差異方面的性能。結果發現，當總體均數差異較小時，Schuirmann無差異檢驗的效能在樣本量較大的情形下會高於student's t檢驗。而總體均數差異較大時，則Schuirmann無差異檢驗的效能總是高於student's t檢驗。

關鍵字

Schuirmann無差異檢驗雙向單側t檢驗 ； student's t檢驗 ； 虛無假設檢驗 ； 樣本容量不等 ； 方差不齊性

並列摘要

It is not uncommon for researchers in psychological and educational fields to evaluate the equivalence of two group means. Equivalence testing approaches (e.g. Schuirmann's test of equivalence) are more appropriate than traditional null hypothesis approaches in these situations. A simulation study was used to compare Schuirmann's test of equivalence with Student's t test for detecting population equivalence under common conditions. Several variables were manipulated in the study: sample sizes, the ratio of sample sizes, population mean configuration and population variances. The result indicated that, when the difference between the means was less, Schuirmann's test of equivalence was more effective than Student's t test at detecting population mean equivalence with large sample sizes; and when the mean difference was larger, Schuirmann's test of equivalence was always a better choice.

並列關鍵字

Schuirmann's test of equivalence two one-sided t test ； Student's t test ； null hypothesis test ； unequal sample sizes ； variance heterogeneity