Let's say I want to measure something inside a population

let's say I want to measure something inside a population

what is the number of people I need to survey ?

bout tree fiddy

30.

All of them.

this, at least

why

statistics

What about t-distribution? ... as a matter of fact how do you even know it's normally distributed in the first place?

Depends on how the thing you're measuring is distributed, if it's normal then you might get away with a couple of dozen people. If it's more esoteric then you'll need more people to get a better idea of the underlying distribution. Either way your guiding principle is to make your sample as large as possible. Also read up about sampling methodologies, to a certain extent they can be more important than the your sample size.

Any number so long as you select them with a reasonably random method from the sampling frame. Obviously the larger your sample size the more representative/accurate your measurements will be, but that is up to you.

Jesus fuck is this what happens when people forget 100 level statistics?

>statistics
not really a science mate

The more answers or possibilities that survey can give the more people you need

as many as you have funding to survey

You don't, and this is why statistics is Psychology-tier.

There are apox 7.1 billion people.
So, about 500 million would give a fairly accurate picture.
It would also weed out poor collection methods pretty quick, because by the time you reach 10 million, if you've been lying about the results, at least a few million would call you out on it.

Statistics are flawed because they're always an attempt at confirmation bias.
This is why they choose small sample sciences.

Here are the "practices" that depend on confirmation bias... I mean statistics:
- Psychology
- Psychiatry
- Sociology
- Civil Engineering
- Medicine
- Law
- Teaching
- Pharmaceuticals
- Etc.

Notice a trend?
It's iatrogenesis in the name of confirmation bias.

>what is the central limit theorem

Put the tin foil hat aside for a second user, the goal of hypothesis testing is to disprove a hypothesis, not confirm it

But some science are purely hypothetical and never move on.
That's an absolute infallible fact.
That is why they're called "theoretical sciences", but they're not theoretical, they're hypothetical because they're against testing.

Your personal attack is meaningless by the way.
History will look back on theoretical sciences and shake its head like it always has.

It's really more circular desu

>observe a thing
>form hypothesis based on current understanding
>test hypothesis
>either refute or fail to refute hypothesis
>update understanding of the thing
>go to step one

The scientific method explicitly includes testing hypotheses

Statistics is just a tool used in this process

>10 different answers in this thread

this is why no one takes statisticians seriously

Well I never thought I'd see someone take tree fiddy as a serious suggestion.

There's only two answers in this thread, one guy gives an equation for n, and another says "as many as you can afford to survey"

For one very simple survey, 10 people let you say things like "3 of 10 people --> 30%"

Doing 3 replications its just the minimal logical amount for making a deduction on science.

>First time you see something y when x happens: cant say shit about it
>Second time you see something y when x happens: Suspect
>Third time you see something y when x happens: possible correlation
>lolfaggotnth time: muh paper

Then a cheeky chinese makes a paper better than you