How to calculate sample size with low event frequency

• Nov 17th 2012, 07:58 AM
battler
How to calculate sample size with low event frequency
How do I calculate the sample size with a really low event frequency.

I have a population of 2^32 with an unknown distribution. But a event frequency of 0,062%. How do I calculate my sample size? I can pick my sample space at random and have no problem with large spaces (~100.000).

I'm new in statistics please excuse my ignorance.
• Nov 17th 2012, 03:48 PM
chiro
Re: How to calculate sample size with low event frequency
Hey battler.

What attribute are you trying to measure? Is it a proportion? Normal mean?

Basically depending on what this is you will need to then pick a significance value to construct a confidence interval and then solve for n so that the interval is smaller in length than the threshold.
• Nov 18th 2012, 04:24 AM
battler
Re: How to calculate sample size with low event frequency
We have a lot of computers in the world. They are being adressed with an IP-address. The total amount of address space is 2^32. However we dont use all this for the internet so roughly it is (2^32)*0.6 . I would like to do some research on a "small" group of computers (lets say 100.000) and than make a estimate for the entire world.
All computers run different kind of services. I want to take my sample group and see who of them are running this service, and that might tell me something about the amount of computers in the world who run this service. I can choose my sample group completely at random. I prior research showed that roughly 0,062% of the computers in there sample group ran this service. Btw this service is completely random some run it, some dont (as stated in the previous research chances of running this service are small). How large does my sample group need to be in order to state something like this about world wide usage.