This Months Haul~August Edition

[Zerodash]

Man, you guys don't know shit about ethics. Something can only be wrong if you're a bad ass for doing it.

What?

[Port]

Why didn't I think of this before?

If you do that, this thread will shrivel down to about 2 posts. (Pre piracy flamewar, anyway)

[Timber]

How does the nielsen ratings system work?

http://www.nielsenmedia.com/whatratingsmean/

I still fail to see how a sample of 5000 households out of 99 million gives an accurate reading.

[kingoffighters]

http://www.nielsenmedia.com/whatratingsmean/

I still fail to see how a sample of 5000 households out of 99 million gives an accurate reading.

It is statistics. It is proven that if you have a sufficiently large sample size, you can limit your error to a fairly small percentage. It is proven mathematically, but it is not intuitive.

[Timber]

It is proven that if you have a sufficiently large sample size, you can limit your error to a fairly small percentage.

I don't know much about statistical math but does 5000 out of 99 million really qualify as a "sufficiently large sample size"?

[kingoffighters]

I don't know much about statistical math but does 5000 out of 99 million really qualify as a "sufficiently large sample size"?

Very nice, you picked up on the 'fuzzy' language that statisticians like to use! Well, there is an exact formula you can use to figure out how large a sample size you need based on what you think the total population is. But I don't remember it hehe and I am too lazy to look for it. If you are really interested, you can take an introductory statistic course and you will learn all that and more. Personally, I think it is pretty interesting stuff and you will understand how statistics can be twisted to represent all kinds of BS

But for now, you can basically take it as faith that Nielsen knows what they are doing

[Diff-chan]

For a more intuitive example, all the polls you see around election time use similarly small samples, and they seem to be fairly accurate.

[Yoshi]

I don't know much about statistical math but does 5000 out of 99 million really qualify as a "sufficiently large sample size"?

Actually yes. Once you cross about 1,000 in total population, the percentage you need to sample drops off significantly. In this case, using simple random sampling and taking a worst case scenario of a 50/50 "success" rate, you would only need 385 data points to achieve a 95% confidence level and a 5% error rate for a population of 99 million.

Aren't you glad you asked?

In case anyone was dying to know, 664 data points gets you to a 99% confidence level.

[Timber]

Aren't you glad you asked?

Actually yes. I kinda get what you're saying. And once I get some extra time & money I may take an introductory statistic course at the community college here. It does seem interesting.

[Yoshi]

Actually yes. I kinda get what you're saying. And once I get some extra time & money I may take an introductory statistic course at the community college here. It does seem interesting.

Believe it or not, I do this kind of stuff every day. It's amazing how few people have even a basic understanding of statistics, which is why TV and newspapers get away with ridiculous analyses without people getting up at arms.