Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Suspicious Discontinuities

danluu.com

Post reply on HN
Print view View original

31–40 of 307 posts

Re: Suspicious Discontinuities

#31

Post by lostmsu »

DavidVoid wrote: post #25
lkbm wrote: post #12

OKCupid has a graph of reported height for men[0]. It's both shifted 2 inches to the right of average, but also slightly flattened at the normal peak of 5'10". It's not nearly as pronounced as the graphs in this post, but it's still notable. [0] https://theblog.okcupid.com/the-big-lies-people-tell-in-onli...

Slightly off-topic but this reminded me of what is perhaps their most depressing pair of graphs. Look at "A woman's age vs. the age of the men sho look best to her" https://i.kinja-img.com/gawker-media/image/upload/1439100617... and then compare that with "A man's age vs. the age of the women who look best to him" https://i.kinja-img.com/gawker-media/image/upload/1439100617...

Why is it depressing? Biologically this must be tuned to the best age for having kids, and men's age hardly matters.

Re: Suspicious Discontinuities

#33

Post by aeturnum »

This is a lovely post and good general demonstration on this phenomenon. The drug charging graph is particularly stark.

I think the most interesting graph is the final one, for marathon finishing times. The idea that people will fudge numbers to meet (or avoid) official definitions is commonly believed (and shown repeatedly here). It's more interesting to observe that it's not just people using motivated reasoning, but also that people will alter their performance to hit highly visible targets.

Re: Suspicious Discontinuities

#35

Post by hammock »

robocat wrote: post #24
lkbm wrote: post #12

OKCupid has a graph of reported height for men[0]. It's both shifted 2 inches to the right of average, but also slightly flattened at the normal peak of 5'10". It's not nearly as pronounced as the graphs in this post, but it's still notable. [0] https://theblog.okcupid.com/the-big-lies-people-tell-in-onli...

Average height changes significantly depending upon your age cohort. If few older people sign up to OKCupid, that would be a sample bias and the average height would be noticeable higher. Edit: “Adjusting for age, adults ages 18-74 years with an annual income of $15,000 and over were observed to be 1.2 inches taller than those with an annual income under $4,000”, amongst other serious biases (whiteness, education) th…

The difference in height for US age cohorts (among adults, obviously) is driven by ethnicity- and younger cohorts are shorter on average.

Re: Suspicious Discontinuities

#36

Post by stared »

Ad Polish end of high school exams: quite a few years ago I created an interactive widget to explore it https://github.com/stared/delab-matury.

There are a few jokes about the Matural Distribution (Matura is the name of this exam) or the Polish Gaussian.

It is the Polish language, which is the most discontinuous. However:

- There are quite a few scores for the essay which are subjective (e.g. on the style).

- It is an official procedure that in the case of results on the edge, the person marking should mark an ambiguous score in the favour of the student.

Re: Suspicious Discontinuities

#37

Post by bradlys »

EpicEng wrote: post #6
eloff wrote: post #3

Earlier quoted context omitted.

I was always small until about age 16, and I always sucked at sports and never really got into them - partly as a result. I wonder if we shouldn't group kids by size for sporting activities instead of by age. There's a famous statistic in Canada that NHL players are much more likely to be born in January and February than September-December. The reason being is they were nearly a year older than other kids they were…

>I wonder if we shouldn't group kids by size for sporting activities instead of by age. Depending on how small your buckets are, you may end up with a scarcity problem (players and admins/coaches). Also, age matters a lot in regards to coordination. A 13 year old who is roughly the size as a ten year old will, on average, have a large advantage. Some people just aren't cut out for athletics genetically. Some grow lat…

> Some people just aren't cut out for athletics genetically. Some grow later on and may or may not be interested. Age seems like a fair sorting criteria and general enough to not introduce too much complexity.

Age isn't particularly fair but it isn't the worst either. As a kid, being around kids younger than you can be really displeasing. When you're a kid, the only thing you want is to be treated like you're older/better than you are. If you keep getting put on teams for people who are small - it's likely you're being put on teams for people who are smaller than you.

I had to play baseball with 7-yr olds for about 3 years. It wore me down as a kid. I gave up on baseball at that point and said I wasn't doing it anymore. The reason I had to play with them is because no coach wanted me on their team because they saw me as too small. (Small town, they didn't group by age as much as they did by size) I'd see kids go up to bigger-kid teams and get bigger but I stayed the same size.

As a kid - it was super disheartening to be stuck with the same 7 to 8 year olds for all that time and seeing people you used to play with in the bigger kid league. It sucked.

That said - baseball is probably the ONLY sport I can imagine being where you could move kids up by age and not by size as much. Football, soccer, basketball, etc. all involve a serious size component to them whereas baseball is less so - in my mind.

For reference - I grew up to be a normal sized adult. I was just super late on the curve compared to my peers.

Re: Suspicious Discontinuities

#39

Post by ehnto »

stared wrote: post #36

Ad Polish end of high school exams: quite a few years ago I created an interactive widget to explore it https://github.com/stared/delab-matury . There are a few jokes about the Matural Distribution (Matura is the name of this exam) or the Polish Gaussian. It is the Polish language, which is the most discontinuous. However: - There are quite a few scores for the essay which are subjective (e.g. on the style). - It is…

> It is an official procedure that in the case of results on the edge, the person marking should mark an ambiguous score in the favour of the student

Given that it's subjective, I find that a really generous and positive take on marking protocol. It takes some perspective to note that while you can't see the value in someone's work, there may still be value in it.

edit: After reading through the article in full, I understand perhaps that's not what's going on exactly. Just teachers trying their hardest to find ways not to fail the exam as it is such an important life step for that person. Still a virtuous thing to do!

Re: Suspicious Discontinuities

#40

Post by ehnto »

lkbm wrote: post #12

OKCupid has a graph of reported height for men[0]. It's both shifted 2 inches to the right of average, but also slightly flattened at the normal peak of 5'10". It's not nearly as pronounced as the graphs in this post, but it's still notable. [0] https://theblog.okcupid.com/the-big-lies-people-tell-in-onli...

Anecdotally I saw that skew in real life before online dating. Women and men would all tell me I looked "About" 5'10", which I decoded as shorthand for "You're definitely not 6 foot"
Post reply on HN