General Discussion
In reply to the discussion: Contrary to popular opinion, DU is supposed to be for politically *liberal* people. Either that or [View all]bemildred
(90,061 posts)And mainly we do those things because it's makes the math convenient and it is fucking great for propaganda purposes, you can make statistics support just about anything.
How do you prove some population follows a normal distribution? You "sample" a "large part" of the population and draw a graph and take the difference integral for the two. If it's small enough, you think it must be normal. You do not assume modest sample sizes will do the job, that would be using statistics to prove statistics applies.
This is almost never done, large samples, it is expensive and the results are hard to pooh-pooh if they come out wrong. Nobody wants that. A 5% confidence level is just right to go either way with, use it if you like it, ignore and keep trying, it if it's not convenient.
But most of the time it is just assumed. The whole point of statistics is to be able to assume that small samples will do, and to make calculations convenient with relatively the small sample sizes so you can attack your enemies with the results. But unfortunately this procedure creates a lot of factoids and pseudo-knowledge.