Analysis of singly ordered two-way contingency tables英文文献资料.docVIP

c JournalofAppliedMathematicsDecisionSciences,4(1),83–98(2000) ReprintsAvailabledirectlyfromtheEditor. PrintedinNewZealand. Analysis of Singly Ordered Two-way Contingency Tables J.C.W. RAYNER john rayner@.au School of Mathematics and Applied Statistics, University of Wollongong, North?elds Avenue, NSW 2522, Australia D.J. BEST john.best@cmis.csiro.au CSIRO Mathematical and Information Sciences, PO Box 52, North Ryde, NSW 1670, Australia Abstract. Pearson’s statistic is investigated for nominal-ordinal two-way contingency tables in whichwewishtotestforidenticalrows. Thestatisticisexpressedasasum,the?rstsummandof whichisastatisticgivenbyYates(1948),andexamineslocatione?ectsforthenominalcategory. Thesecondandsubsequentsummandsre?ectthecorrespondingmoments: forexample,dispersion, skewness, kurtosis etc. The summands are shown to be weakly optimal in that they are score statistics. Keywords: Categoricaldata,logisticmodel,log-linearmodel,orthonormalpolynomial,two-way data. 1. Introduction In the social sciences and in more specialized areas such as sensory evaluation, it is common to obtain categorized ratings for a number of items. For example, if eyesightisunderconsideration, onemighthave?vecategoriesofeyesightranging frompoor toexcellent,withobservationsonbothmalesandfemales,givingatwo by?vecontingencytableofresponses. Asoneofthecategorizationsisordered,it is possible to do a more thorough analysis than that given by the usual X2 test P for a two-way contingency table. Sometimes, although the X2 test may not be P signi?cant,ane?ectmaybesuggestedbyoneofanumberofanalysessuggestedin theliterature,including (i) giving the categories equi-spaced scores and using a regression analysis as in Yates(1948); (ii) using nonequi-spaced scores based on mid-rank values as in Bross (1958), Conover(1998,p.281)andNair(1986); (iii) linearlogisticmodelsasinMcCullagh(1980); (iv) log-linearmodelsanduserde?nedassignedscoresasinAgresti(1984,p.84); (v) thecumulativechi-square