How is that a bias? It's just basic logic, you can't justify a confidence in something if you don't understand how it can be proven or disproved.brimstoneSalad wrote:Perhaps you should not: there's a bias in that people have misplaced confidence in things they think they understand better.

Well, yes. If other galaxies weren't visible with normal telescopes, the Big Bang Theory wouldn't have been postulated.brimstoneSalad wrote:You mean unobservable due to redshift or what?

I meant to say that saying "Things in languages are right or wrong only because people who speak those languages agree on that." is wrong for the same reason saying "Money has value because people agree it has value." is wrong.brimstoneSalad wrote:Not sure what your point is.

But it doesn't matter if they made new observations or have simply done new statistical analyses on previous observations. Had they made 20 new experiments, each with the p-value of 5%, chances are, they would have probably also "found" a type of jellybean that caused acne. Doing the same experiment again and again is also a form of p-hacking.brimstoneSalad wrote:I already talked about that xkcd comic that showed the principle...

How is it not low, it's less than 1/1000?brimstoneSalad wrote:What are the odds of taking ten red samples from an infinite pool of 50%-50% red and green?

Actually not that low.

So, we can say "At least 75% of all apples are red." with 94% certainty, right?brimstoneSalad wrote:If there are even just 75% red apples, you get .75^10 which is almost a 6% chance.

With the Havlik's Law, the calculation is not that simple. What's the probability that a word in a modern Slavic language appears to follow the Havlik's Law if it doesn't actually do that? Well, almost every word in Proto-Slavic had a yer in it (almost every noun ended with a yer, which disappeared in modern Slavic languages), and we can simplify that every yer was either vocalized or unvocalized in some modern Slavic language (let's ignore the clusters yer+r and yer+l, which obeyed different laws in quite a few languages, and let's ignore that, had the Havlik's Law been invalid, we wouldn't expect all the vocalized yers to turn into one sound in each modern Slavic language). So, we can say that the probability of a word appearing to follow the Havlik's Law when it doesn't actually do that is 0.5^number_of_ProtoSlavic_yers_in_it. Let's say there are 500 words in each Slavic language (there are probably more) to test for the Havlik's Law. I don't know right now what number should be put as the average number of yers in a Proto-Slavic word, but even if we put only 1.0, the p-value is still insanely low. Even if we try to calculate the p-value for the Havlik's Law being apparently accurate for 300/500 words (using the binomial coefficients), it's so low the calculator can't calculate it. The p-value gets to 5% already at 268/500 words, so if the Havlik's law was apparently only 53% accurate, and I guarantee you it's more accurate than that.

Yes, but that's only because we know a lot about apples. If we didn't, it would be scientific to assume the sample is indeed random, right?brimstoneSalad wrote:That sample is likely not very random at all.