|Abstract: ||在處理列聯表(Contingency tables)資料時,適合度檢定(Goodness-of-fit-test)的結果如果是顯著的話,通常意味著配適模式的不恰當,而這其中一個可能原因是資料中存在離群細格(Outlier cells)。因此我們常希望能夠針對問題,找出離群細格。離群細格的偵測通常藉由不同定義的殘差作為工具,從而衍生出各種不同的偵測方法。不過以往文獻中的探討多局限於二維列聯表離群細格的偵測,對於三維或三維以上列聯表的情形則沒有比較具體的討論。在這個研究計畫中,我們依據Brown(1974)以及Simonoff(1988)的建議,提出一個適用於三維列聯表離群細格偵測的方法。模擬實驗的結果顯示,在絕大多數的情形下,這個方法的表現較Brown(1974), Simonoff(1988)及統計套裝軟體BMDP的4F程序有顯著的改善。|
Chi-square goodness-of-fit tests are usually employed to test whether a model fits a contingency table well. When the test is significant, we would generally like to identify the sources that cause significance. The existence of outlying cells that contribute heavily to the test statistic may be one of the reasons. Many procedures have been proposed to locate outlying cells in a two-way table. Generalization to higher-dimensional tables, however, was either unclear or never mentioned. Motivated by Brown (1974), and Simonoff (1988), we propose an alternative method which is applicable to any three-way table. Based on simulation results, we find that the procedure performs reasonably well, and it has better performances than Brown's, Simonoff's, and BMDP program 4F most of the time.