Current Events > This exchange is great

(+) Yeah! (+)

https://gamefaqs.gamespot.com/a/user_image/4/4/4/AAZt-XAAEUec.jpg
https://twitter.com/instrumenthull/status/1639723003104706561?t=oL3JMVJTU-HIntPQfDzFYQ&s=19

Lmao.

I can't imagine being publically wrong about such a simple concept for over 3 years.

I don't get it.

AldousIsDead posted...

I don't get it.

TC doesn't have to explain why this is funny because if he did Elon would make him pay

AldousIsDead posted...

I don't get it.

One guy says that if you have enough numbers you don't need to bother with randomization. The other is making fun of him.

Dude you gotta give some context to this you might as well have posted a story in Gaelic.

Now TC is just being stupid

That's no hydroxyl ion, that's my wife!

AldousIsDead posted...

I don't get it.

In *the majority* of cases, the only way to estimate a causal effect of a treatment is to manipulate the treatment directly through random assignment. The one dude says "i just need a large number of subjects" and has been holding onto this position for 3 years, publically arguing with academic statisticians.

In stats 101, you might have heard the example of ice cream sales being correlated to violence. You can collect all the data you want, and you will aways find that relationship. But it is not causal and is driven by a 3rd variable, heat. That's the key thing the dude is missing. You can't just measure, because there are many many observed and unobserved confounding variables. And the only way to gaurantee success is to randomly assign the treatment, as if you do this with enough subjects, all other factors will average out.

[LFAQs-redacted-quote]

Is it safe to assume some people simply have a richer mental life?

Anyway, this is just a dude very confidently stating "correlation is equal to causation as long as i have a high enough number of observations".

ellis123 posted...

One guy says that if you have enough numbers you don't need to bother with randomization. The other is making fun of him.

And the big enough numbers guy is wrong? I assume they're talking about programming applications, so while on the face of it big numbers guy seems wrong, but I don't know enough about the topic to judge.

AldousIsDead posted...

And the big enough numbers guy is wrong? I assume they're talking about programming applications, so while on the face of it big numbers guy seems wrong, but I don't know enough about the topic to judge.

Correct.

Though apparently I was wrong in how I read it as I don't get COVxy's correlation:causation take at all. I guess the complete lack of context got me as well.

hmmm yes quite shallow and pedantic indeed

[LFAQs-redacted-quote]


NoMeLx22x posted...

Dude you gotta give some context to this you might as well have posted a story in Gaelic.

AldousIsDead posted...

I don't get it.

Scientific experiments need participants. "n" = a participant or subject being studied. Let's say you want to study if vaping causes cancer. You would do this by finding a bunch of people and then splitting them into two groups: one group will not vape (the 'control group') and one group will vape (the 'treatment group').

But hold your horses. What if these two groups you make aren't similar? There could be a 'selection bias' introduced when you sort the groups. Maybe the treatment group has more family histories of cancer. This will bias the results of the experiment, making them unreliable.

To fix this, scientists use randomization. They make sure that who gets into the experiment and what group they're in is not affected by any other variable or human error.

The size of 'n' has nothing to do with whether your control and treatment group are sufficiently similar, or whether your sample of participants is reflective of the entire population. Well technically, '30' is an important number where certain statistical formulas hit a limit, and introducing a bigger n after 30 will not make your hypothesis test's error range any smaller. But this aspect of 'n' is separate from the problem of randomization for predictive experiments.

Why are people on ce so averse to clicking things?

Posting the context of a twitter thread is fucking nonsense and would require like 15 screenshots or links to tweets. Meanwhile you can just click and scroll around, or complain.

COVxy posted...

Why are people on ce so averse to clicking things?

Posting the context of a twitter thread is fucking nonsense and would require like 15 screenshots or links to tweets. Meanwhile you can just click and scroll around, or complain.

Because no one wants to click on Twitter.

COVxy posted...

Why are people on ce so averse to clicking things?

I dont click tweets and nowhere was there indicated that further context would be provided if I did.

BurmesePenguin posted...

I dont click tweets and nowhere was there indicated that further context would be provided if I did.

Sounds like a you problem lol.

Quite not.

Lmao stay within the bounds of possibility COV

You gotta many problems but the future ain't one

Cringe