Student 09 posted on Thursday, August 26, 2010 - 5:56 am
I wonder whether and how in Mplus a variable measured on the within-level can be used as grandmean-centred variable in a twolevel regression model.
Suppose my major interest is in the effect of school-type - a between-group variable - on students' math achievement as mesaured at the within-group level, and I also want to control for the infuence of students' SES as measured on the whithin-level. Such a situation would seem well-suited to apply grandmean-centering of students' SES. Introducing the grandmean centered variable in the equation would yield results controlled for SES, given that my major interest is on the effect of school type.
However, the Mplus default is obvioulsy a (latent) group-mean centerinng approach, where student's SES would be decomposed in a within- and between part.
Contrary to this default, I don't want to decompose SES in a within and between part but only treat it as grandmean-centred predictor.
Would that be possible in Mplus, despite the required %within% & %between% parts of the model?