Date: Tue, 21 Sep 1999 17:34:40 +0200
Reply-To: Massimiliano Marinucci <mmarinucci@INNER.ES>
Sender: "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From: Massimiliano Marinucci <mmarinucci@INNER.ES>
Organization: INNER Startegic Market Research
Subject: Re: quick-cluster
Content-Type: text/plain; charset=iso-8859-1
Dear Arnaldo,
what do you mean for significant results? Recall that Qcluster is a
nonhierarchical clustering method which need to have an a priori
definition of the number of clusters, therefore you should have any a
priori hypotesis on the clusters you get.
Anyway, to check if your solution is reasonable you should examine the
associated ANOVA table which SPSS gives in Output. This is useful in order
to see if there are differences in the means between the groups you
computed. Next you should examine differences among the cluster centers
(centroids) and compare it with other solution you get with a higher or
lower number of clusters. Of course, you may also save each individual
cluster membership and cross tabulate it with sociodemographic data, in
order to identify their sociodemographic structure.
Hope this helps.
Ciao
Max
****************************
Massimiliano Marinucci
Junior Researcher
INNER Strategic Market Research
c/Velazquez 109 6ēD
28006 Madrid
Tel. +34 91 5631128
Fax. +34 91 5626813
e-mail:mmarinucci@inner.es
http://www.inner.es
****************************
ISPO wrote:
> Dear List,
> Here is my problem.
>
> How can I know if a quick-cluster (k-means) gives significant results?
> I use SPSS 6.0 .
>
> TIA
>
> Arnaldo.
>
> ::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
> :: ::
> :: Istituto per gli Studi sulla Pubblica Opinione ::
> :: ::
> :: Via Toti 2, 20123 Milano, Italia (U.E.) ::
> :: ::
> :: tel. 39 2 48004104 - fax 39 2 48193197 - email ispo@it.net ::
> :: ::
> ::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
--
ISPO wrote:
> Dear List,
> Here is my problem.
>
> How can I know if a quick-cluster (k-means) gives significant results?
> I use SPSS 6.0 .
>
> TIA
>
> Arnaldo.
>
> ::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
> :: ::
> :: Istituto per gli Studi sulla Pubblica Opinione ::
> :: ::
> :: Via Toti 2, 20123 Milano, Italia (U.E.) ::
> :: ::
> :: tel. 39 2 48004104 - fax 39 2 48193197 - email ispo@it.net ::
> :: ::
> ::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
| 
