Use of factor/object scores in Probit

[Andrew Robinson-12]

I am working on a national survey and using categorical principle
components analysis (CATPCA) as a data reduction technique and using the
saved object scores in subsequent probit estimates. To take account of the
sample design I would normally apply the appropriate sample weights to any
estimation as I am interested in the population parameters. However, in
the above two-stage scenerio I am unsure whether the CATPCA analysis
should be carried out on weighted or unweighted data since the object
scores will also be weighted in the probit equations.

Any guidance or references to related work would be appreciated.

Thanks

Andrew