Quote Originally Posted by LilimoLimomo View Post
I think you may be getting something confused, because when it comes to extrapolating data, that is exactly what it means. You get a large enough sample of users, get their data, and then provided that there aren't any confounding variables, you can safely assume that the data you found proportionately generalizes to the overall population. That's how researchers get statistical data for everything.

And in this case, we have a really great sample size of over 100,000 players. The call to action is put out, and of those players, less than 400 care enough to express negative feedback. If we generalize this data to the overall player population, those 400 players represent more than just 400 players. But the same is true for the 999,600+ players who weren't bothered enough to provide negative feedback; they too represent more players. Which is why the percentage of 0.4% generalizes to the rest of the player population.
This is completely wrong and I hope you never talk about statistics ever again.

By your insane logic, homelessness is not a problem that affects women because only a few homeless people spoke out about the problem on NPR. Opioid addiction clearly does not harm millions of Americans because only dozens are talking about it in the New York Times.

By your insane logic the COVID survey that the government sent out shows that only tens of thousands of Americans are negatively affected by COVID since they bothered to fill out the opt-in survey.

You don't even understand what confounding variables mean. Trying to establish causality is problematic if there are confounding variables that are unaddressed. Trying to find proportions is NOT affected by confounding variables. It is affected by sample bias or attrition which are NOT what any statistician calls confounding variables.

The forums are a biased representation of the entire playerbase since generally the more active and invested players will participate. That does NOT imply that the rest of the playerbase are completely neutral. The way to address sample bias is not to conclude that everyone out-of-sample behave the exact opposite of what people in-sample are like. That is ludicrously stupid.

I would hope you would take the L but you're likely not to so please realize you're talking to an actual statistician.