I am working as a researcher in genetics field. I am proposing a study that I need to provide the appropriate sample size to get statistical approval.

The expected frequency of our event will be something between 8% and 12.8% When I used the following to calculate the sample size using this site, http://www.nss.gov.au/nss/home.nsf/pages/Sample+size+calculator, the confidence interval was 0.024, proportion was 0.104, confidence level 95%. The result is 622.

The issue now, i want to propose only 300. How i could argument this statistically?

NB. I have very basic stat knowledge. Can someone please explain in plain terms I could understand?


closed as unclear what you're asking by Juho Kokkala, Sycorax, gung, Silverfish, usεr11852 Feb 19 '16 at 21:42

Please clarify your specific problem or add additional details to highlight exactly what you need. As it's currently written, it’s hard to tell exactly what you're asking. See the How to Ask page for help clarifying this question. If this question can be reworded to fit the rules in the help center, please edit the question.

  • 1
    $\begingroup$ What about the power in this sample size? Power is the probability that a test correctly rejects the null hypothesis? I wouldn't ignore that in your evaluation. That said, what if you plug in the result of an n=300 and work backwards to determine what the resulting power, CI, CL and proportions are? $\endgroup$ – Mike Hunter Feb 19 '16 at 16:24
  • 4
    $\begingroup$ How do you know you want to propose sample size $300$ without knowing the reasons? $\endgroup$ – Juho Kokkala Feb 19 '16 at 17:04
  • 1
    $\begingroup$ What you did was calculate the size needed to find a 0.104 proportion to within +/- 0.024, in 95% of samples. That might be more precision than you need. The calculator says that a sample size of 300 will give you +/- 0.035 in 95% of samples. It's not completely clear that you used the sample-size calculator in the way you intended. Typically one wants to distinguish, say, the proportions in 2 groups, which is a different power calculation. If you are doing genetics, you either need a strong statistical background or need to work with someone who does. $\endgroup$ – EdM Feb 19 '16 at 17:33
  • $\begingroup$ The budget of this study is good for 300 but not for more. $\endgroup$ – user1502182 Feb 19 '16 at 17:33
  • $\begingroup$ I will try to work backward to get the Cl and CL. It is a good idea $\endgroup$ – user1502182 Feb 19 '16 at 17:34

Larger sample is always better statistically, but sometime it's too expensive. So, your argument would be outside usual statistical arguments. You'd have to argue that weighing in the cost of obtaining the sample, the expected cost of error etc. the optimal sample size is smaller than suggested by pure statistical argument.

For instance, at 5% significance you expect certain cost of error with going with null vs. alternative given the costs of obtaining the sample.


Not the answer you're looking for? Browse other questions tagged or ask your own question.