When I talk about statistics on HN it feels like I'm taking crazy pills. The HN community pretends to be intellectual, but the level of mathematical education here is terrible.
> Moreover, you're hyperbolizing yourself. Namely, you criticize the title, which is not the author's title at all. Their letter is titled "Report From Working Group on the Role of Standardized Test Scores in Undergraduate Admissions"
The authors clearly say "SAT and ACT scores are highly predictive of academic performance at Dartmouth". That is what they want to show.
And this is totally false. They cannot accurately predict high vs low academic performance based on SAT or ACT scores. Their data clearly shows this is impossible.
> Except if you read the text, that's not at all what the authors claim. The authors do not claim the title (again... that's HN's title, not the paper's). The authors say:
Huh?! That's literally their claim.
They even say in the conclusion "They are significantly predictive of academic success at Dartmouth"
This is just totally wrong. They are a statistically significant but practically insignificant predictor of academic success when you define it in the most absurd and irrelevant way (first year GPA).
> It does... in fact... mean that there is a linear relationship in the observed data with the given levels of significance.
Summarizing data with some line, without any statistics, and saying "look the line slopes up", does not mean there is a linear relationship. At all. You cannot conclude one from the other.
The fact that this isn't obvious is truly frightening.
You're right those sentences are there. My bad, I should have kept a closer eye. I looked at the author's title, which is indeed not as sensational as the HN title though.
> They are a statistically significant but practically insignificant predictor of academic success when you define it in the most absurd and irrelevant way (first year GPA).
As I said in my post, the bulk of their argument references other studies at similar institutions
> Summarizing data with some line, without any statistics, and saying "look the line slopes up", does not mean there is a linear relationship. At all. You cannot conclude one from the other.
They run their own linear regression. In no way is the argument "look the lines look linear".
> Moreover, you're hyperbolizing yourself. Namely, you criticize the title, which is not the author's title at all. Their letter is titled "Report From Working Group on the Role of Standardized Test Scores in Undergraduate Admissions"
The authors clearly say "SAT and ACT scores are highly predictive of academic performance at Dartmouth". That is what they want to show.
And this is totally false. They cannot accurately predict high vs low academic performance based on SAT or ACT scores. Their data clearly shows this is impossible.
> Except if you read the text, that's not at all what the authors claim. The authors do not claim the title (again... that's HN's title, not the paper's). The authors say:
Huh?! That's literally their claim.
They even say in the conclusion "They are significantly predictive of academic success at Dartmouth"
This is just totally wrong. They are a statistically significant but practically insignificant predictor of academic success when you define it in the most absurd and irrelevant way (first year GPA).
> It does... in fact... mean that there is a linear relationship in the observed data with the given levels of significance.
Summarizing data with some line, without any statistics, and saying "look the line slopes up", does not mean there is a linear relationship. At all. You cannot conclude one from the other.
The fact that this isn't obvious is truly frightening.