Re: Regarding SPSS discussion

[Art Kendall]

David was correct.  This is a bear of a problem and highly specialized.

Your data is spatially  and time auto correlated.  That means the data are doubly repeated measures. You have data that is structured by place and date.
However, there are methods to take advantage of this. If ignored, all kinds of problems can result.

You need to be very explicit.  Although synoptic is a general English word derived from Greek, it most likely has a special meaning in meteorology.

I am not familiar with using any form of factor analysis for doubly repeated measures data.  For finding dimensions, you may want to consider your data to be summarized in a 3 mode individual difference multidimensional scaling (INDSCAL). This uses a stack of distance/similarity/correlation matrices. I have only worked with 2 mode INDSCAL.  Maybe some form of structural equations modeling. 
 
I suggest that you
1) write a clearer description of your data.
2) describe what the profile of a cluster might look like. Are you looking for different profiles across places on the same date? Profiles that are different by time of year and by place so that a profile could occur in Place A in summer and Place B in winter?, etc.
3) give some links to results that look like what you are trying to achieve

After list members see that info and understand the problem, we may be able to suggest how to get where you want to go or recommend lists that would deal more directly the suggested approaches.

Over the next few weeks I'll only have intermittent access to the net, but others on the list may be able to help you get a clearer description of your needs.
 
Art Kendall
Social Research Consultants

On 6/25/2011 10:00 AM, Zeeshan Ali wrote:

Sorry Art, I forgot to mention link. Here it is;

http://spssx-discussion.1045642.n5.nabble.com/PCA-R-Matrix-Determinant-0-and-quot-not-positive-Definite-quot-td4512844.html

Peace.

On Sat, Jun 25, 2011 at 7:10 PM, Art Kendall <[hidden email]> wrote:

Which forum are you interested in.  I do not see your name on SPSSX-L
http://listserv.uga.edu/archives/spssx-l.html
or on
<a moz-do-not-send="true" href="news://news.eternal-september.org/comp.soft-sys.stat.spss" target="_blank">news://news.eternal-september.org/comp.soft-sys.stat.spss

Are you thinking of another forum? did you use another name?  what was the subject?

Art

On 6/25/2011 2:11 AM, Zeeshan Ali wrote:

Dear Art

Please spare sometime to have a look on the on going discussion on the forum SPSS. I have tried to explain the data set i am working on. 

Peace.

Muhammad Zeeshan

===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD