Stats question...

[AdamH]

How do you determine whether something is statistically significant? I took many a stats course but in trying to help my co-worker in Fundraising I'm getting stuck on how to answer her question:

She is doing a "prospect" mailing, which means there are two test groups, A and B. A will receive a direct mail package featuring a celebrity, B will receive a package with an international development theme.

The general rule is that you use at least 40 000 new people for a prospect mailing to get signficant results. We only have 34000. The question is...versus 40K can 34K names still produce a statistically significant result?

[Velvet]

9 times outta 10, yeah.

[Mr. Musicface]

Hey ahess. I only have a very small amount of stats/survey knowledge, but my understanding of how "significant" a result is relate to what you are asking, how random your sample is, how big a population your sample is meant to represent, etc. So it's gonna vary based on exactly what you're testing for. For instance, polling groups for elections often use relatively small samples (a few thousand people usually I believe) to represent how a whole country is gonna vote, but they're really careful to have a varied group and ask very precise questions.

Hey I don't know if this helps but I stumbled on this calculator:

http://www.surveysystem.com/sscalc.htm

There's some links to definitions and such as well.

Good luck!

- M.

[purple foot]

its all about the chi square or something.