P-value is not a criterion for deciding whether a hypothesis is true

P-value is not a criterion for deciding whether a hypothesis is true.

Other urls found in this thread:

en.m.wikipedia.org/wiki/Sunrise_problem
twitter.com/SFWRedditGifs

yes it is

HOW WILL OP EVER RECOVER?/

Could someone explain to me what calculation that one math guy used to calculate the probability the sun will rise tomorrow, given it's risen for the past 10,000 years? I've always been very interested in that after reading it in a book, but never could find the calculation.

Everyone knows Bayesian math works given enough input. The problem that no one agrees on is the priori assignment. I always wondered why they didn't simply calculate two results: one with a priori of 0% and one with a priori of 100%. As more information was added these two values would converge, being an upper and lower bound.

I recently read an article about some hugely intelligent and influential computer science mind realised that the brain doesn't work probabilistically and we instead think causally. STATISTICS IS A LIE

You sound like a fucking mongoloid. Nobody knows how the brain works, so there hasn't been """"""some hugely intelligent and influential computer science mind"""""" who """"""realized"""""" how the brain works.

Veeky Forums is for science and math, not literary interpretations of science of math.
The Kolmogorov axioms are all you ever need to know about probability, and every other candidate interpretation is sound if and only if it is structurally homomorphic to a model of these axioms.

Baww go somewhere else then. Veeky Forums is shit, you can't fix it to your liking.

It certainly is a criterion for deciding which papers to publish. A shitty feminist 'study' on the relationship between sexist icecubes and toxic masculinity can fish up a juicy p value and the study is published based on that one number being proof that the hypothesis is true.

No shit we don't think probabilistically. Any Psych. 101 student could have told you that.

>one with a priori of 0% and one with a priori of 100%. As more information was added these two values would converge, being an upper and lower bound.
in practice, bayesian modeling is not done with a single probability value. instead, they use a distribution, usually one of the more well characterized ones that behave nicely wrt expectations and variance. beta priors are pretty common, and there are parameterizations of the beta distribution that give you some different profiles, including one set that gives you a uniform distribution

furthermore, you wouldn't test just one prior distribution, you'd try a couple that make sense given your knowledge of the subject

says the undergrad who still believes truth is defined mathematically

Correct, it is a criterion for deciding whether a hypothesis is false.

Rejection of the null does not imply that the alternative is true.

Good thing no one decides whether a hypothesis is true based on a single study

>computer science
>intelligence

pick one

Let f be a non-negative real-valued function on {0,1} such that f(0)+f(1)=1. What is f(1)?

Bayesian: well, without something else to go on, I'd guess f(1)=1/2

Frequentist: I don't know, but if I have two integers n and N, and if I assume n/N is approximately f(1), then I would guess f(1) is approximately n/N.

Mathematician: You're retarded.

en.m.wikipedia.org/wiki/Sunrise_problem

maybe this?

You're very right. Fisher addressed this at some length in several articles. It was his intention that p-values should be interpreted only as (circumstantial) evidence against the null hypothesis, and never as evidence for a competing hypothesis.

A fund manager returns >10% for 4 straight years.

Frequentist approach: Invest all your money in the fund manager

Dank approach: investing all your money in a 10x leveraged ETF

high beta low vol switching is the future. just you wait

It is. Name a model that isnt truw via math.

P-value does not determine whether a hypothesis is true, but that doesn't mean they aren't useful. They allow you to reject the null within some acceptable type 1 error rate.
The Bayesian approach isn't always better either. Suppose a drug with potentially severe side effects shows stupendous effectiveness in treating a particular disease in phase 1 and phase 2 trials, but then shows no significant effectiveness in a phase 3 trial. Should the effectiveness demonstrated in the phase 1 or phase 2 study be used to "update" the results of the phase 3 study and conclude the drug is effective enough for market?

The truth value of a mathematical statement depends on the model it is considered in. In Euclidean geometry, a triangle's angles total 180 degrees, but their sum can be more or less depending on the model of geometry used (spherical or hyperbolic, resp.). Only a statement true in all possible models could be an absolute truth. Otherwise it is relatively true.