If your descriptive variables are continuous, you should look at a
discriminant function analysis. In SPSS, you simply add the new cases
to the file the clustering was done on. They will have missing values
on the variable for cluster membership. In a DFA, the first phase will
produce the profiles of the groups. {the "tests" are not valid as such,
but can be used as a sort of relative measure.) The classification phase
of the discriminant will tell which cluster each of the new cases is
closest and give you the probability of membership in that cluster and
the probability that a member of this cluster would be so far from the
centroid.
If you have mixed levels of measurement, the ANSWERTREE module in SPSS
would be one place to start.
Hope this helps.
Art
[EMAIL PROTECTED]
Social Research Consultants
University Park, MD USA
(301) 864-5570
Joe wrote:
> Hi, all,
>
> I have several clusters. And now I need to find out the common
> characteristic within each group, so that these common characteristic
> can be used as criteria to group the new data.
>
> Could someboday tell me which statistical method I should use?
>
> Thank you very much.
>
> Joe
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
. http://jse.stat.ncsu.edu/ .
=================================================================
