/sci/ - Science & Math » Thread #12218892

248KiB, 625x350, avvo.png

View Same Google iqdb SauceNAO

Could you explain to me Fequentists vs Bayesian?

Anonymous Sun 11 Oct 11:51:12 2020 No.12218892 View Reply Original Report

Quoted By: >>12219272 >>12219484

I am gonna learn statistics at school but never understood the difference between frequentists and baesyan. If you should explain it to me?
As far I know frequence is ''how much x times a thing happen in a given timeframe''
while Bayesian is incorporating a prior an always changing the outcomes based on that. What's the frequentist approach?

Anonymous

Anonymous Sun 11 Oct 2020 14:55:05 No.12219272 Report

Quoted By: >>12219528

>>12218892
Even position of atom in unverse can be non continous function, why would frequency be possible? You have to do furier transfor to understand statistics... There are some high frequency, som low frequency repetitions.

Anonymous

Anonymous Sun 11 Oct 2020 16:41:48 No.12219484 Report

Quoted By: >>12219532

>>12218892
Frequentist: the percentage of the event in a lot of trials, e.g. coin flips
Bayesian: how new data changes your expectation about an event, which you can’t repeat a million times like the coin flips, e.g. elections

Anonymous

Anonymous Sun 11 Oct 2020 17:00:47 No.12219528 Report

Quoted By:

>>12219272
>You have to do furier transfor
fourier transform what?

Anonymous

Anonymous Sun 11 Oct 2020 17:02:43 No.12219532 Report

Quoted By: >>12219602

>>12219484
let's assume I would apply bayesian to a coin flip. How would I do? After some tosses with tail (let's say 3) I say ''it's more likley to have tail then an head'' ?

Anonymous

Anonymous Sun 11 Oct 2020 17:12:50 No.12219568 Report

Quoted By:

Frequentist va Bayesian is a question of philosophy, really. In the Frequentist view, everything is a posthoc observation on large trials. You're observing the truth as it was using just the data, and nothing about your own belief ever enters the picture.

The Bayesian view is that you say, "hang on, we have perfectly good tools right here in the math for representing human belief, we just have to decide what statistical model most accurately represents that belief". The Bayesian view is applied in the middle of the analysis, not as a data autopsy. You are updating your model of the data as you go.

Ultimately, neither is "right". They're two interpretations of the same mathematical processes. They should get to the same answer in the end. If you are an accomplished statistician, you might prefer to use one mode of thinking, but you should be able to apply both schools of thought throughout your work as appropriate.

Anonymous

Anonymous Sun 11 Oct 2020 17:25:07 No.12219602 Report

Quoted By:

>>12219532
In general, with bayesian analysis your hope is always that someone has done the leg work for you and your data can be modeled using exact closed form solutions. For example, if your prior and posterior both end up being normal, there are simple expressions for the posterior mean and variance. If you dont have a simple case, you have to do some more complicated computation.

In your case, you want a beta distribution at your prior and posterior. I believe there are closed form solutions for beta posteriors but they're not as clean to look at as they are for normals. See here for some visual examples https://towardsdatascience.com/visualizing-beta-distribution-7391c18031f1

Capcode	All Only User Posts Only Moderator Posts Only Admin Posts Only Developer Posts
Show Posts	All Only With Images Only Without Images
Deleted Posts	All Only Deleted Posts Only Non-Deleted Posts
Ghost Posts	All Only Ghost Posts Only Non-Ghost Posts
Post Type	All Only Sticky Threads Only Opening Posts Only Reply Posts
Results	All Grouped By Threads
Order	Latest Posts First Oldest Posts First

Your latest searches