I would have thought Answer Tree would also do the trick - that is, to yield a set of rules that describes each group generated by cluster analysis. I believe each of the four algorithms in Answer Tree respects ordinal predictor variables.

Regards,

Jason.

PS In addition to French, Answer Tree is available in English too!

> -----Original Message----- > From: SPSSX(r) Discussion [mailto:SPSSX-L@LISTSERV.UGA.EDU]On Behalf Of > Census - Data Analysis & Data Mining > Sent: Saturday, 4 March 2000 1:08 > To: SPSSX-L@LISTSERV.UGA.EDU > Subject: Re: Cathegories that characterize groups > > > The concept of "valeurs-test" are used in french packages to identify the > categories of variables that best characterize groups generated, for > example, by a cluster analysis. > > You will find references about this concept in the french software SPAD > (Système Portable pour l'Analyse des Donées) developed by CISIA > (www.cisia.com). > > In the documentation of module "DEMOD" you will find a complete > description > of the "valeurs-test" and another related indicators to characterize > groups. > > As SPSS does not have this feature, when I have to characterize > clusters or > some partition of the data set by one o more nominal variables, I export > the data matrix to SPAD to obtend a rapid and detailed description of the > characterization, identifying the variable categories that best describe > each cluster. > > Even hough SPSS is very powerfull, when I have to work with nominal or > ordinal variables, I found SPAD better (specially than Categories module). > > I prefer to work with SPAD, specialy to perform "dimension reduction" via > correspondence analysis (binary or multiple) followed by a classification > with groups characterization. > > However, I am working in a SPSS script to reproduce the groups > characterization algorithm. > > Some books references are: > > SPAD reference manuals > > Cahiers de l'Analyses des Données (As I know, the characterization > algorithm is described in some volume of this publication). > > Lebart L., Morineau A., Tabard N. (1977) Techniques de la Description > Statistique, Méthodes et Logiciels pour l'Analyse des Grands Tableux - > Dunod - Paris > > Lebart L., Morineau A., Fenelon J.P. (1979) Traitement des Données > Statistiques, Méthodes et Programmes - Dunod - Paris > > Benzecri, J.P. - L'Analyse des Données - Tome 1: La Taxinomie, Tome 2: > L'Analyse des Correspondances > > Hope this help you. > > Néstor Martínez > Statistician > Census - Data Analysis & Data Mining > Buenos Aires > Argentina > > census@dynamo.com.ar > > ---------- > > De: Nichols, David <nichols@SPSS.COM> > > A: SPSSX-L@LISTSERV.UGA.EDU > > Asunto: Re: Cathegories that characterize groups > > Fecha: Viernes 3 de Marzo de 2000 5:27 PM > > > > SPSS doesn't offer anything called a valeurs test. It's > impossible to say > > how the CROSSTABS residuals would relate to the test, since I'm not sure > > what the test does. It sounds like the test is a single or > global one for > a > > predictor, whereas the residuals are one per cell in the > crosstabulation. > > Certainly, a variable that would predict a grouping would have to have > some > > adjusted residuals that are large, but I don't recall seeing that one > could > > order predictors on such a basis. > > > > David Nichols > > Principal Support Statistician and > > Manager of Statistical Support > > SPSS Inc. > > > > > -----Original Message----- > > > From: Chorianopoulos Antonis > > > [mailto:Antonis.Chorianopoulos@EKEBI.CULTURE.GR] > > > Sent: Monday, February 28, 2000 7:14 AM > > > To: SPSSX-L@LISTSERV.UGA.EDU > > > Subject: [SPSSX-L] Cathegories that characterize groups > > > > > > > > > I've asked a few days ago about a test that is used to > > > describe a group of respondents (ie respondents belonging in > > > a certain cluster that has derived after a cluster analysis) > > > by finding categories of other nominal variables that > > > characterize them. > > > Actually the test i'm referring to is a test called > > > "valeurs-test" and used mostly in France. I'm afraid > > > the only things i know about it is that it finds the > > > categories that characterize our group of interest and ranks > > > them according to their importance. > > > > > > I was wondering if the Adjusted Standardized Residuals in the > > > Crosstabs procedure can help us > > > do something like that (you look for categories with Adj. > > > Standardized Residuals outside (-2, 2) to characterize your group). > > > > > > Thank you very much for the help > > > > > > > > > > > > Antonis.Chorianopoulos@ekebi.culture.gr > > > >