Hacker News Reader

Comment by BoiledCabbage | Hacker News Reader

>... but these effects are definitely exaggerated by the reweighing of population proportions.

Thanks for sharing the data, but your own data doesn't appear to support that comment.

If I'm reading correctly, the overall drop in US scores was 18 points. The drop in US scores of the two largest populations were 16 points and 17 points respectively. This isn't Sampson's paradox at all. This is exactly what it looks like, each subpopulatuon dropped significantly in score between the pre-covid test and this one.

To call a difference in score of 17 vs 18 (the overall average) an "exaggeration due to reweighting" is quite a misleading stretch of interpretation.

View on HN

Yep, overall drop in US scores is 18 and whites dropped by 16, so really the majority of the drop is the fact that all students dropped.

Definition of exaggerate:

To consider, represent, or cause to appear as larger, more important, or more extreme than is actually the case; overstate.

As you point out, the weighted difference is ~17 which you can clearly see in the data. Increasing this by a full point is quite significant and most people do not even have the intuition that it is even mathematically possible to increase 17 to 18. This is why we teach Simpson's paradox. I also posted the data so people could come to their own conclusion. Just because you use a definition of exaggeration to mean some nonsensically large difference, that does not make my claim misleading.

I would say that it’s quite disturbing that anti-immigrant bigotry, draped in the most threadbare example of pseudo-intellectualism, is the most upvoted comment in this thread, but this unfortunately seems to be the order of the day. Perhaps as a corollary to this study, someone should do a statistical analysis of what has caused this decline in reasoning ability among tech workers.

This is what you get if you shun uncomfortable discussions, try to shirk finding explanations in problematic cultural practices and promote right wing idiocy by running around with a half collapsed air-castle of comfy half-truths, kept propped up only screaming power of the civilized half of society.

Controversial statement I fully believe in. It is not a decline. Tech workers were always like that. We just liked to not talk about it and call everyone who notice "sjw". Because it felt shameful.

> I would say that it’s quite disturbing that anti-immigrant bigotry

At most hypothesis may be false, or unsupported by the evidence, but never is a hypothesis by itself "bigotry".

A quick review of the parent comment would tell you that it never rises to the level of being a hypothesis because it is rooted in category errors and false distinctions that make it neither testable or falsifiable. For example, the appeal to comparison of immigrants as a bloc with shifting American fictions about what constitutes membership in its racial categories. These boundaries are too ill-defined and overlapping with each other to be used in this fashion. What is the process by which one transitions from being an immigrant to one of these groups and how would you observe and control for how any of these (unstable) requirements are correlated to test performance? And of course the same appeal Simpson’s paradox would also have to be applied to the category of immigrant itself before it could ever be considered in this way, since this umbrella contains even more and diverse members than the racial categories with which it is contrasted.

TLDR: Not all suppositions are hypotheses.

That's nonsense, it's clearly (obviously!) possible that the math score of a country decreases because of immigrants having lower test scores and immigrant share increasing. Which makes it a valid hypothesis. Any problems with definitions are orthogonal to that.

If you believe definitions constituent to a hypothesis are ever orthogonal to it, this would be a mark in the ledger in favor of there being an overall decline in reasoning ability. By this logic, I can simply say “x is a responsible for a decline in y, as evident by the comparable measure of z” without ever having to define any of those variables and always be correct. Does that seem rational or useful to you?

Look, I can understand what you say, and you can understand me, without us defining all or any of the words we use. We don't need a definition of definition. We can know what things mean without having a definition of "knowledge" or "meaning".

That more abstract claim was never what is at dispute. The OP’s claim instead requires clear and consistent definitions of the categories involved in order to be a falsifiable hypothesis, which were not provided. Your defense of this is apparently that because it also exists on the low end of the spectrum of some basic intelligibility that also makes it a hypothesis, which is, again, also incorrect. It’s a supposition that never rises to this level because it lacks these aspects.

Reply on news.ycombinator.com