Generator Microsoft Word 11 (filtered medium) Dear all,
I want to use a weighting variable in SAS. In the case of independent observations and the model
yi = alfa + ei
I can use PROC GLM and specify a weight matrix W and then the resulting estimate of alfa is
(X' W X)-1 (X' W y)
i.e. sum[wi*yi] / sum[wi], where i=1, 2, . , n.
In a mixed model, yij = alfa + ai + eij, according to Crowder&Hand, 1990 (p 89) I can estimate alfa as
(inverse{sum[X' S-1 X]}) * sum[X' S-1 y],
where S-1 is the inverse of the covariance matrix and the summation is made over the n independent subjects.
What is the expression for the estimate of alfa in a mixed model if I want to include a weighting matrix W?
Best regards,
Eva Andersson
Occupational and Environmental Medicine
Sahlgrenska University Hospital, Goteborg University
Goteborg
Sweden
|