In previous releases, if you specified a WEIGHT statement and one or
more of the multiple comparisons options, PROC GLM estimated the variance
of the difference between weighted group means for group i and j
as
-
MSE×([1/(ni)] + [1/(nj)])
where MSE is the (weighted) mean square for error and ni is the
size of group i. This variance is involved in all of
the multiple comparison methods. Beginning with Release 6.12, the variance
of the difference between weighted group means for group i and j
is computed as
-
MSE×([1/(wi)] + [1/(wj)])
where wi is the sum of the weights for the observations in group
i.
Copyright © 1999 by SAS Institute Inc., Cary, NC, USA. All rights reserved.