Date:         Fri, 16 Dec 2005 03:59:30 -0500
Reply-To:     Sto <Malte.Stopsack@UNI-GREIFSWALD.DE>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:         Sto <Malte.Stopsack@UNI-GREIFSWALD.DE>
Subject:      Mixed models for (dependent) data?
Content-Type: text/plain; charset=ISO-8859-1

Hello,

we have a little problem with modelling with dependent data:

In our study we have several families with their children. We computed an
ANOVA (four groups) and a multivariate GLM (four groups + covariates) for
the children. Because there are sometimes (not every time) siblings in the
data set from the same family the values of these children aren’t
independent.

What we did so far: We analyzed the data for all children, for one
children per family (randomly selected) and for the mean values for the
sibling of a family. The problem with the first solution is the non-
independence of the data, the second and third loose power.

But someone told me that it is possible to handle this problem with mixed
models without the mentioned limitations but I haven’t found a practical
solution which satisfies me in SPSS until now. (Another possibility could
be bootstrapping?)

Any advice or information would be greatly appreciated.

Best regards

M. Stopsack

PS:  SPSS-Version 12.0, in some days 14.