Login required to started new threads

Login required to post replies

Any math/stats wiz's on here. Simple statistics question
Quote | Reply
Let's say I have a population of 5000 people of which the individuals answer a question with a 'yes' or 'no'. I don't have the time to look at all the answers to find out how many yes's and no's there are. Does anyone have the formula's or a reference web page they can direct me to that would provide the formula's for the following:

What sample size is required to have 95% confidence that the yes's are +- 5% of ____%?

I will alter the parameters to get different deviations and confidence intervals.

Much much appreciated. I need this for validating data in a target database but it has been so long since I have had to do these calculations that I can't remember and I have had no luck finding this on the internet.

________________
Adrian in Vancouver
Quote Reply
Re: Any math/stats wiz's on here. Simple statistics question [AJHull] [ In reply to ]
Quote | Reply
You want a +/- 5% margin of error? Try this site

http://americanresearchgroup.com/moe.html

You can directly calculate it, but you can play with the sample size until you're close enough

_______________________________________________
Quote Reply
Re: Any math/stats wiz's on here. Simple statistics question [AJHull] [ In reply to ]
Quote | Reply
http://en.wikipedia.org/wiki/Margin_of_error

Has the actual formual

_______________________________________________
Quote Reply
Re: Any math/stats wiz's on here. Simple statistics question [jhc] [ In reply to ]
Quote | Reply
Thanks John. This is exactly what I am looking for. How does this work for boolean values though? This appears to be for a set of values. A bit more info on what I am doing...

This is part of a migration project. I acquired 10,000 records from the source system and I have imported them into the target system. I have a few people assigned to verifying the data in the target system to confirm that the data in there and imported properly. So my plan is to have them take a sample of the records and have them investigate and come back with the results that (for instance) 98.3% of the records are successful. How can I translate this to say something about the expected results of the whole population?

________________
Adrian in Vancouver
Quote Reply
Re: Any math/stats wiz's on here. Simple statistics question [AJHull] [ In reply to ]
Quote | Reply
I think I have made sense of the best approach based on your second post. Thanks

________________
Adrian in Vancouver
Quote Reply
Re: Any math/stats wiz's on here. Simple statistics question [AJHull] [ In reply to ]
Quote | Reply
Cool - I came up with a sample size of 370 (out of total 10,000) for a 5% margin of error. Does that sound like what you got?

_______________________________________________
Quote Reply
Re: Any math/stats wiz's on here. Simple statistics question [AJHull] [ In reply to ]
Quote | Reply
I forget the exact way to do this, but the second post is assuming an essentially infinite sample population. That is not your case. Since you will not be replacing the samples after you take them, the approach listed there will be a conservative estimate.
Quote Reply
Re: Any math/stats wiz's on here. Simple statistics question [jhc] [ In reply to ]
Quote | Reply
That's what I got. Thanks John.

________________
Adrian in Vancouver
Quote Reply