Reread what you just wrote. Statistical theory
always holds in real life by its very nature.
If you can disprove the central limit theorem, then you just invalidated modern statistics. I doubt you can.
I am sure that virtually every single social statistician will be shocked by how fallacious their practice is.
It is obvious that you have no knowledge of statistical practices whatsoever. I suggest you take an introductory course at the very least.
You're right, I only know the basics. I'll know what you're talking about in detail later this year.
What I do know, though, is that most rules in statistics are well and good in theory, but when applying them in real life, you get an utterly different result.
If you try rolling a six-sided die six times, I can almost guarantee you won't get one of each side as a result.
In that sense, statistics aren't infallable.
Like flipping a coin, the more occurences you have in a sample, the more your result will change.
I don't care about what a statistician will say. I'm right because it makes sense. Think for yourself. Roll a dice or whatever to prove it, if you don't understand or agree.
Comparing two items of differing sample space is misleading because of the above reasons. If you were to change the sample space, then the results would definitely change.
I also know that if you're trying to find a trend, the more data you use, the more accurate your averaged result will be.
There are more cars, so the mean of accidents per unit of time will be more accurate. There are less powerplants, so the mean of accidents per unit of time will be less accurate.
I haven't been taught statistics in detail, but what I've said is common sense. Take it or leave it. If you still disagree, I just suggest you re-read what I've said with an open mind.
It is statistically valid to to make inferences from data with different alpha levels, we must do it, because we don't get idealized data.
Normally, I'd agree with you. But since there is a massive difference between the amount of people driving a car, and the amount of nuclear power stations, I'd have to disagree on that note. The effect on the result would be apparent with
that much difference. If the sample spaces were realatively close to one another in size, then I wouldn't have a problem.