http://en.statistik-tutorial.de/tutorials/weighting-spss.html

*(Q)        My sample has 40% males and 60% females.
        My population has 49% males and 51% female.
        I would like to automatically weight my sample cases in order to
        approximate my population.

*(A)  Posted to SPSSX-L list by Raynald Levesque on 2002/09/13.
        Note that dispersion and confidence interval statistics produced by
        SPSS are not valid for a sample weighted like this.



DATA LIST FREE /sex.
BEGIN DATA
1 2 1 2 2
END DATA.
LIST.
SORT CASES BY sex.
SAVE OUTFILE='c:\temp\original data.sav'.


AGGREGATE OUTFILE=*
        /PRESORTED
        /BREAK=sex
        /wsex=N.
FLIP.

SELECT IF RTRIM(case_lbl)="WSEX".
COMPUTE popmale=49.
COMPUTE popfem=51.

COMPUTE sampmale=100*var001/(var001 + var002).
COMPUTE sampfem =100*var002/(var001 + var002).

COMPUTE wgtmale=popmale/sampmale.
COMPUTE wgtfem =popfem /sampfem.

FLIP VARIABLES=wgtmale wgtfem  .
COMPUTE sex=2.
IF case_lbl="WGTMALE" sex=1.

MATCH FILES TABLE=*
        /FILE='c:\temp\original data.sav'
        /BY=sex
        /DROP=case_lbl.
FORMATS wsex(F5.4).
WEIGHT BY wsex.

非常感谢楼上的热心解答，我仔细研究下！

非常感谢楼上解答

在SPSS Modeler里加入“平衡节点”很方便的调整。