Hello,
I would like to use the REGWGT Subcommand in the Mixed model module
to compensates for violation of the homoscedasticity assumption by
weighting cases differentially. I know the independent variable which
causes heteroscedasticity but I do not know how to estimate the
regression weights for correlated data sets. I tried the weighted
least squares (WLS) weight estimation to estimate my weights. I
assumed that if my model is correctly specified I do not have to worry
(too much) about repeated measurements since I am not interested in
inference and regression weight estimates should be unbiased. I used
the weights in my mixed model and the problem of heteroscedasticity of
my residuals disappeared.
However, I read in the SPSS help that the regression weights will be
applied only to the covariance matrix of the residual term. Now I am
not sure anymore if my approach is still valid or at least okay?
Best wishes, Felix
--
GMX FreeMail: 1 GB Postfach, 5 E-Mail-Adressen, 10 Free SMS.
Alle Infos und kostenlose Anmeldung:
http://www.gmx.net/de/go/freemail
Locked
