Question

Topic: Research/Metrics

Sampling

Posted by Anonymous on 2/1/2006 at 11:41 AM ET 100 Points
Ok.. this is my last try to get answer to my puzzle. Given that my population is 750 and i want margin of error to be +/-5% at the confidence level of 95% what is the smallest sample size that I can hope to get. I need the smallest possible sample, yet, statistically significant. Thanks everyone for your help!
To continue reading this question and the solution, sign up ... it's free!
SUBSCRIBE
Or, SIGN IN

RESPONSES

  • Posted on 2/1/2006 at 11:48 AM Accepted
    254 is the sample size that will give you +/-5% at the 95% confidence level.

    That being said - you can have a smaller sample and still have statistically valid results. The smallest sample size you can have to statistically valid results is 30. With a sample of 30 the central limit theorem kicks in. So what you are trading of or gaining is margin of error. For a sample of 30 the margin of error is roughly +/-18%.

    A margin of error of 5% is the standard for market research. However, often you need to look at is the cost to obtain that margin of error worth it? What is the consequence of making the wrong decision?

    This is not an easy concept to graps so feel free to contact me directly for more information.
  • Posted by wnelson on 2/1/2006 at 1:01 PM Member
    OK, remember in the last question I replied that there are four parameters to sample size. One of those is the probability of outcome and that the largest sample size is when the probability of "satisfied" versus "unsatisfied" is 50%. This is where the answer of 254 comes from. However, two questions ago, you said that you felt that your outcome was going to be 80% "satisfied." If this is true, then you can survey just 186 people and your error will be +/-5% to a 95% confidence. HOWEVER: If you find that the survey results show just 60% or 40% (both are equal) satisfied, your error goes up to 6.13%. If it's 50%, the error goes up to 6.25%.

    If you are willing to live with 10% error, you can survey 86 max (50% "satisfied") or just 57 if your assumption of 80% satisfied is valid.

    If the confidence is not a problem, you can go to 90% confidence and get a sample size of 200 with an error of 5% with a response of 50% satisfied or 141 with a "satisfied" response by 80%.

    There's no magic "sample size reduction" pills here. Four parameters to play with that determine sample size:
    1) Probability of the outcome (given a question, the number of people likely to answer "yes" versus "no."
    2) The population size
    3) The error - the +/- 5%
    4) Desired confidence - how sure you are that the results will be within the given error

    You can't change population size. If you increase any of the other parameters, your sample size goes down. Play with the calculator:

    https://www.berrie.dds.nl/calcss.htm

    You can find the appropriate trade off in sample size to precision/accuracy/confidence.

    I hope this helps.

    Wayde
BACK TO TOP

Post a Comment

Related Discussions

Community Info

Top 25 Experts
(Research/Metrics)