How do you do a weighted least squares regression to correct for homoskedasticity?

#1 SPSS.

I have an ethnicity variable as an IV and BMI as a DV, with unequal ethnic group sizes. I ran an ANOVA for BMI and included the Levene's test, which was significant, indicating hetereogeneity of variance. I would like to correct for heteroskedasticity using a weighted least squares regression. However, I am completely baffled by the procedure. In SPSS, under weight estimation (the WLS regression), you have to select three variables: IV, DV, and "weight variable." Which one is which--I only have two variables, ethnicity (IV) and BMI, which was my DV and which I gather is the variable that suffers from heteroskedasticity. I also don't know what to do with the "power options" at the bottom. If anyone knows how to do this procedure and could help me, I'd be very grateful.

Edited to say that in my title, I meant to say correct for heteroskedasticity--I want to increase homoskedasticity. Major duh. :>