A question on statistics and errors

[Streetwalker]

Thank you Pappy.

I am not saying that the technique is not use, but I have no idea where the idea came from

The theory of probabilities  , Bliase Pascal 

[Borchester]

Try this:- How to Analyse Survey Data: Methods and Examples | SurveyMonkey

Thank you Pappy.

I am not saying that the technique is not use, but I have no idea where the idea came from

[Unlucky4Sum]

 . . . .I was right, it is a huge simplification, you have to take into account the desired confidence level and whether it is or is not a so called Standard Distribution

See https://openstax.org/books/introductory-statistics/pages/8-formula-review  

[Unlucky4Sum]

Let us assume that you are conducting a poll on behalf of the Free Beer Party.

Your sample is 1600.

So to find the error bound you find the square root of the sample (which is 40), express that as a percentage of the sample ( [40/1600] x 100 =2.5%) and there you are. You have the error which is +/- 2.5%

Easy peasy. In passing it is why pollsters don't bother with larger numbers. If the sample was 10,000 you would still only have an error bound of +/- 1%.

But my question is, where does the rule come from? I have browsed the internet and even offered mathematically inclined friends ready cash for an answer but to no avail.

Does anyone know?

Looks to me like that's a very simplistic formula that might work in some circumstances but generally won't.  For a start the error bound would have to have some link to the confidence level and it shows none.

But now you've tweaked my curiosity  . . . . .. . 

[Streetwalker]

Let us assume that you are conducting a poll on behalf of the Free Beer Party.

Your sample is 1600.

So to find the error bound you find the square root of the sample (which is 40), express that as a percentage of the sample ( [40/1600] x 100 =2.5%) and there you are. You have the error which is +/- 2.5%

Easy peasy. In passing it is why pollsters don't bother with larger numbers. If the sample was 10,000 you would still only have an error bound of +/- 1%.

But my question is, where does the rule come from? I have browsed the internet and even offered mathematically inclined friends ready cash for an answer but to no avail.

Does anyone know?

Its the rule of Brexit ,the closer the poll is to 50/50 the bigger the margin of error 

[papasmurf]

Let us assume that you are conducting a poll on behalf of the Free Beer Party.

Your sample is 1600.

So to find the error bound you find the square root of the sample (which is 40), express that as a percentage of the sample ( [40/1600] x 100 =2.5%) and there you are. You have the error which is +/- 2.5%

Easy peasy. In passing it is why pollsters don't bother with larger numbers. If the sample was 10,000 you would still only have an error bound of +/- 1%.

But my question is, where does the rule come from? I have browsed the internet and even offered mathematically inclined friends ready cash for an answer but to no avail.

Does anyone know?

Try this:- How to Analyse Survey Data: Methods and Examples | SurveyMonkey

[Borchester]

Let us assume that you are conducting a poll on behalf of the Free Beer Party.

Your sample is 1600.

So to find the error bound you find the square root of the sample (which is 40), express that as a percentage of the sample ( [40/1600] x 100 =2.5%) and there you are. You have the error which is +/- 2.5%

Easy peasy. In passing it is why pollsters don't bother with larger numbers. If the sample was 10,000 you would still only have an error bound of +/- 1%.

But my question is, where does the rule come from? I have browsed the internet and even offered mathematically inclined friends ready cash for an answer but to no avail.

Does anyone know?