Well, that’s the set up of the problem. It doesn’t really change the point if everyone has a 1/10 or 1/100 chance to commit a crime on a randomly chosen person, it’s just easier to state and write out if it’s 1.
1 Like
Well I’m going to start a Twitter account where I begin every tweet with “holy shit” and then say something that some Conservative likely won’t get in trouble for. I’ll rack up millions of followers.
The fatal flaw is the assumption that crime victims are randomly selected from the national population.
It’s been clear for a while that Elon’s aims with Twitter are not related to making money. He’s burned billions of dollars doing whatever it is he’s doing. Scratching a $25M check to Tucker because he thinks it is epic is nothing at all to him.
3 Likes
Of course they aren’t, but the correct point that I think you were trying to make was that one population being smaller doesn’t prove that the number of crimes against the other population should be expected to be larger all else being equal - which is what normalization twitter was suggesting.
Does he have an audience though? Bill O’Reilly had a huge audience, got canned, and disappeared into the ether and he’s way more charismatic than Tucker.
I can never comprehend this stuff until I put a ratio against it. $25M to a $100B-aire is like someone worth $1M paying… $250
Elon doesn’t give a shit
3 Likes
The correct point I was making is that the statistical validity or invalidity of Elon’s tweet had nothing to do with what’s wrong with it. You wouldn’t change your mind about it if a more nuanced statistical analysis proved that black people really were disproportionately likely to commit cross-racial crimes. Nor should you.
I didn’t really intend for this to be an argument, but you said “shoddy statistical analysis” in regards to the revulsion that was being backfilled. Now you’re saying it wasn’t about shoddy statistical analysis. So what was shoddy in the analysis that was backfilling the revulsion?
My favorite way to visualize numbers so big they become meaningless:
How long is a million seconds? Just over 11 days.
How long is a billion seconds? Over 31 YEARS.
3 Likes
No, the analysis is very shoddy because criminals don’t victimize a random person from the national population.
(((Zuckerberg)))
imagine a society where a guy who reads laura loomer tweets and believes them has 100 billion dollars
15 Likes
I don’t know what normalization twitter is. Do you mean Kareem Carr? I don’t see where he said that we should expect one result or another in his original thread. He was discussing potential biases and made some suggestions.
https://twitter.com/kareem_carr/status/1654862484359004162?s=20
I think “Is any relevant data missing?” is particularly important here and the biggest reason the chart Elon responded to is bad. There’s a famous example of this from aerospace, but maybe Elon isn’t familiar with it; that would be unsurprising. I will post about it later.
I don’t know but I think @bobman0330’s problem is with me rather than anything said on twitter. I think he’s mad because I say mean things about Elon.
Maybe he’s saying my analysis is bad because I suggested starting by considering what the chart would look like if crime was random. I’m not a statistician so maybe that is bad but it’s just a first step. Assuming I can trust the source, next I’d compare it to the actual data to see how big the difference is.* From there, there would be questions about what caused the difference. I might not get very far answering those. Oh well.
*But we are here. And it’s pretty clear to me that the original chart was presented in a deliberately biased way to make the difference from random look as big as possible.
fuckin’ hell man
The WW2 airplane drawing with the missing bullet holes often attributed to Abraham Wald is a more well-known example of selection bias but I had in mind what happened in a late-night meeting between NASA managers and Morton Thiokol engineers before Challenger was launched. The engineers knew it was risky but they couldn’t make their case.
People were acting under pressure and they were looking at the first plot below. The relationship between o-ring behavior and temperature is unclear because data is missing. If they had considered the second plot, maybe things would have been different. Source.
Kareem Carr redid Elon’s plot just to check the numbers.
https://twitter.com/kareem_carr/status/1655614490124357632?s=20
And then he added back the missing data. Maybe this somewhat less impressive version would have gotten the same response from Elon, IDK.
https://twitter.com/kareem_carr/status/1655614530163204097?s=20