中華心理學刊 民 69, 22卷, 一期, 13-21
Chinese Journal of Psychology 1980, Vol. 22, No.1, 13-21

吳瑞屯（國立臺灣大學）

摘要

    本文介紹一個適用於各種型態因子設計變異分析程式的基本運算邏輯。主要的部分是一個全交叉設計變異分析程式。文中探討了两種利用全交叉設計變異分析程式的方法，用以處理巢套設計或混合設計；並根據此等方法設計一個適用於各種等格或不等格因子設計變異分折的通用程式，只要輸入因子數目、名稱及各因子間的關係，並界定隨機效應因子，以及選擇不等格設計時的處理方法 （等比例設計或不加權平均法），該程式將輸出最後的變異分析表，包括必要的 F 或 quasi-F 值及對應的P(H0）值。這個程式最重要的特色在能够適用於混合設計或因子間關係非常複雜的設計，例如多個受試者間變項之間存在有更高階的巢套關係等。

A General Purpose Anova Algorithm for Equal-And Unequal-Cell Factorial Designs

Jei-Tun Wu(National Taiwan University)

Abstract

A general purpose ANOVA algorithm which could apply to all types of factorial designs is elaborated in this paper. This algorithm could equally apply to both equal- and unequal-cell factorial designs. The proposed computer program has the following advantages over the prevailing ones:
(a) a maximum capacity for 15 factors, (b) capable of dealing with mixed designs or very complicated factorial designs, such as a nest relationship between two between-subjects factors, (c) proportional unequal-cell designs and unweighted-means problems, (d) capable of handling 1 or 2 random-effect factors, and (e) cell means and complete ANOVA table being well arranged on the printouts.