The eduspere is all aflutter with NCLB's undercounted minority fiasco. The tension is between statistical signficance on one hand and accountability for traditionally underperforming minority groups on the other. I think I have a solution.
When a school doesn't have enough of one or more minority groups to form a statistically significant group -- just aggregate them. We'll call this group the traditionally underperforming subgroup. The group would include african-americans, hispanics, and native-americans.
If you couldn't form a statistically significant group out of these subgroups, you could then disaggregate the non-northeast-asians from the asian subgroup and include them too. And, if you still needed more in the group, you could include the economically disadvantaged kids to capture some of the lower-performing whites too. That should do it.
Actually, why don't we just get rid of these silly racial subgroups altogether and focus on the real problem area -- the low performers -- the kids who fall on the left half of the IQ curve, no matter what subgroup they come from. These are the kids who need to most educational attention and resources.
We don't need to separately track the high performing african-americans, hispanics, and native americans disproportionately located in the more affluent schools, they're going to perform more like their white and asian counterparts. Likewise, shouldn't we be more concerned about the performance of the low performing whites and asians? They're more likely to peform like the lower performing african-americans, hispanics, and native americans, so let's give them the same attention.
Of course, then we'd have to face the ugly reality of IQ disparities and racial differences, but if we really are concerned about not leaving any child behind, we're going to have to face up to this reality sooner rather than later.