I wouldn't. Sometimes you simply can't draw conclusions (one way or the other) from the available data.
Then strangely, you do exactly that two paragraphs later:
Also, at least for math, you don't really need to look that hard for explanations. US high school math is a disaster area; it is no secret that there's too much focus on teaching rote procedures [1], which will hurt you on PISA...
So making a crude attempt to control for demographics is invalid, and you can't draw conclusions from the data. Yet if you make no attempt to control demographics, it's somehow valid? Huh?
Also, your theory that teaching rote procedures causes poor performance is already refuted by Sanandaji's article. It works just fine for Asian Americans, who perform just as well as wealthy Asians in Asia. Doesn't that piece of data contradict your theory?
> Then strangely, you do exactly that two paragraphs later:
No. I'm drawing conclusions from different things, such as curriculum content, plus the huge body of existing research on teaching math in the US [1].
My larger point is that it's pointless to treat statistics as a black box and try and make conjectures (or worse, massage the data to fit your conjectures), when there are known documented deficiencies.
> Also, your theory that teaching rote procedures causes poor performance is already refuted by Sanandaji's article. It works just fine for Asian Americans, who perform just as well as wealthy Asians in Asia. Doesn't that piece of data contradict your theory?
No. First, it merely shows that when you select the highest performing group, you'll get a higher than average performance. The thing is, the same happens elsewhere, too. For example, the best-performing cantons in Switzerland (a country that is as ethnically diverse as the US [2]) had an average PISA score in math of around 560 in 2009, when the average was 531.
PISA scores are spread out over a pretty wide range. Poor teaching does not make you unable to perform at high levels, it merely handicaps you. Other factors can compensate for it, such as the known higher work ethic of Asian Americans and cultural valuation of intelligence and cognitive skills [3].
Second, he uses a composite score, while I'm specifically talking about math, where the US score is really bad (as opposed to, say, reading).
If it's a disaster, why does the same curriculum work well for Asian Americans (raising their scores to east Asian levels), but badly for white Americans?
First, it merely shows that when you select the highest performing group, you'll get a higher than average performance.
Ok. What causes the variance between high and low performing groups? Since the school system seems the same, it doesn't seem to be educational techniques.
Other factors can compensate for it, such as the known higher work ethic of Asian Americans and cultural valuation of intelligence and cognitive skills
Correct me if I'm wrong, but you seem to be more or less agreeing with me here. Some trait which is very common among Asians appears to be the primary driver of test score deltas.
Then strangely, you do exactly that two paragraphs later:
Also, at least for math, you don't really need to look that hard for explanations. US high school math is a disaster area; it is no secret that there's too much focus on teaching rote procedures [1], which will hurt you on PISA...
So making a crude attempt to control for demographics is invalid, and you can't draw conclusions from the data. Yet if you make no attempt to control demographics, it's somehow valid? Huh?
Also, your theory that teaching rote procedures causes poor performance is already refuted by Sanandaji's article. It works just fine for Asian Americans, who perform just as well as wealthy Asians in Asia. Doesn't that piece of data contradict your theory?