How statistically significant would a random sample of 6000 students be? Random seems like a bad idea for such a relatively small sample. (compared to the rest of the world.)
if it's a true random sample, it should be perfectly good. it might be too small to subdivide into all of the subsamples that they want to evaluate (high-income, well-educated puerto ricans in new jersey, etc.) but as an overall sample it should be more than sufficient. nothing is more powerful than random. (it almost certainly isn't a truly random sample, but perhaps it's close enough).
goddamnit, mcat, now you have me curious why they didn't use pisa results from the us, if those data are available.