Two (hopefully) simple questions. I've taken Statistics in school however I am an intern at a very small non profit and I just want to cover my bases.
1. If we were to put out a survey testing the internet usage of our subscribers (people who have signed up to get our publications offline), would there be an inherent bias if we were to email them this survey?
2. Because our population (total number of subscribers) is 237 (pretty small), I am stuck between deciding to:
i. Send the survey to all of our subscribers (census), that way with an estimated participation rate of 50%, we would have 118 respondents. This would enable us to obtain subgroup analysis (ie. do our 60 year olds have the capability to view our publication online?).
ii. Send the survey to a sample size of n'
where n' = n / [ 1 + (n / N) ]
where n' = sample size corrected for small population size
where n = sample size as per original formula
where N = size of the population
However, when I compute this formula I get:
n' = 153 respondents
(with MoE = +/- 4% and level of confidence = 90%)
Now with a participation rate of 50% (pretty high), I need a population of something bigger than 306, which I just don't have.
So should I do the census? Or just pick an arbitrary sample size, (for cost reasons) that is still large enough to perform some subgroup analysis?
Hopefully, this I haven't made this confusing.
ANY HELP is much appreciated.
The Non-Profit Intern
Last edited by stefan_p; 05-27-2009 at 12:21 PM.
Advertise on Talk Stats