Comment by thaumasiotes | Hacker News Reader

Comment by thaumasiotes | Hacker News Reader

Parent Full thread

Come on. There are no sources that would report this. It makes as much sense as claiming that "as everyone knows, the greatest contributor to tidal pull is the odor of the moon as its cheese rots, which attracts plankton on the ocean's surface". And it has as much scientific backing. Where in the heck are you getting your "facts" from?

Here, from a book chapter subtitled "The case for eliminating the SAT and ACT at the University of California":

    Table 1.1. Correlation of socioeconomic factors with SAT/ACT scores and High School GPA
    +--------------------+---------------+------------------------------------------------+
    |                    | Family Income | Parents' Education | Underrepresented Minority |
    +--------------------+---------------+------------------------------------------------+
    |    SAT/ACT scores  |      0.36     |        0.45        |          -0.38            |
    +--------------------+---------------+------------------------------------------------+

Of the three factors they bothered to look at, family income had the lowest correlation, just below a poor approximation of race and very far below parental education.

( https://cshe.berkeley.edu/sites/default/files/geiser_chapter... )

I looked at that table and immediately thought, "Those are pretty solid correlation values", so I was confused by the point you were making.

I also read the page in the PDF you linked that was the source of the table and became more confused, because it very clearly writes out that:

>What this means is that 40% of the variation in students’ SAT/ACT scores is attributable to differences in socioeconomic circumstance.

I think you are either:

1. Misunderstanding correlation coefficients. Correlation coefficients range from -1 to +1, implying a 100% negative or positive correlation. 0.4 and -0.4 are moderately positive and negative correlation values. That table is saying that if you are an underrepresented minority, you are expected to do 40% worse on the SAT/ACT.

2. Quibbling over the "thing that correlates highest with SAT score" claim, because you feel like 0.4 is not high enough of a number. The only factors with higher correlation values than 0.4 in SAT/ACT scores would be stuff like "Was the student currently injured?" or "Did the student make it to the testing center?"

> I looked at that table and immediately thought, "Those are pretty solid correlation values", so I was confused by the point you were making.

They said "highest", not "solid". Family income isn't the highest factor there.

Of course if having a brain would be on the table there would be no contest. Being the highest correlation in a rather arbitrary list is not a valid qualifier for being a good correlation.

> Of course if having a brain would be on the table there would be no contest.

Is that really what you think? In reality, the possession of a brain can't have any correlation with SAT scores at all, because there is zero variance in the trait.

> That table is saying that if you are an underrepresented minority, you are expected to do 40% worse on the SAT/ACT.

This is separate from the main point, but "underrepresented" minorities are defined by their lower performance on the SAT, so 0.4 isn't impressive at all for that cell.

Also, you've made a gross mathematical error; "you are expected to do 40% worse on the test" is a statement about the effect size, not the correlation.

And parental education, race, and household income are also all correlated with one another... And the paper you cited makes the point that the correlation between socioeconomic background and test scores has increased substantially since 1995. This is not the gotcha that you think it is.

> And parental education, race, and household income are also all correlated with one another...

That's true.

> This is not the gotcha that you think it is.

But your point makes the guy I'm responding to look much worse. After controlling for parental education, the effect of parental income is radically diminished. Most of that 0.36 correlation is just confounding. Parental education isn't even a great indicator itself; it's a proxy for intelligence. It's just a much better proxy than parental income is.

Reply on news.ycombinator.com