Date:         Tue, 15 Jun 1999 10:42:22 -0500
Reply-To:     "Nichols, David" <nichols@SPSS.COM>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@UGA.CC.UGA.EDU>
From:         "Nichols, David" <nichols@SPSS.COM>
Subject:      Re: Help for variance components model?

It sounds like you need to consult with a statistician familiar with your
subject area to find out what the appropriate model is for your design. I'll
be glad to help with the SPSS specifications for a given model, but I have
to know what that model is.
 
David Nichols
Principal Support Statistician and
Manager of Statistical Support
SPSS Inc.
 
> -----Original Message-----
> From: Leena Lankoski [mailto:leena.lankoski@SOIKIO.FI]
> Sent: Tuesday, June 15, 1999 3:29 AM
> To: SPSSX-L@UGA.CC.UGA.EDU
> Subject: Help for variance components model?
>
>
> Dear all,
>
> I am a new subscriber to this list (to any list, in fact) and
> wanted to see whether I could find some help with my problem
> in this way.
>
> I need to run a variance components analysis with SPSS 8.0
> but I am not quite sure how to build the appropriate model
> (i.e. what to put after the DESIGN command), even after
> consulting the manuals and various textbooks. These are the facts:
>
> - The data include one observation from each combination of
> plant/pollutant/year. (There are empty cells because all
> combinations are not possible.) In addition, each plant is
> classified into a detailed industry sector, which in turn are
> classified into crude industry sectors.
> - The question to be answered is how much of the variation in
> the dependent variable is due to industry effects (measured
> at a crude or detailed level), plant effects, pollutant
> effects, year effects, or their combinations.
> - It is my understanding that plant is nested within detailed
> industry classification, which in turn is nested within crude
> industry classification. It is also my understanding that at
> least the year variable ought to be crossed with all others,
> to divide the effects into stable and fluctuating components.
> - Another problem is that some of the variables have many
> levels, and therefore I often run into problems of "integer
> overflow", "not enough memory", or "too many columns". I
> wonder whether it would be possible to somehow split the
> analysis to make it more manageable for the program?
>
> Thank you very much in advance for any help.
>
> Leena Lankoski
> e-mail leena.lankoski@soikio.fi
>