3

Say I'm designing a retrospective study (where I'd collect data from various schools for the last 5 years) to see if consuming milk can contribute to childhood obesity. How do I decide on how many schools to visit? If I remember my stats correctly, in general, the goal is to achieve 95% confidence in our results, right? So how many schools would I have to collect data from so that I will have 95% confidence in the outcome of the study?

Prabhu
  • 131
  • 1
  • 2
  • 1
    Related Question – Rory M Feb 15 '13 at 23:41
  • this is a question of statistical experiment design. The statistical power of the experiment must include all the factors you are controlling for - age, geographical region, family income etc etc. So this really depends. not sure this can be answered here without a lot more information. – shigeta Feb 16 '13 at 00:32
  • 2
    You might consider posting this on Cross Validated, but as shigeta mentioned, you'd need to provide more details. – blep Feb 16 '13 at 09:15
  • Thanks...actually that's the kind of information I was hoping to hear. The age is 5-10 years, geographical region is US, and family income is irrelevant (randomized). – Prabhu Feb 16 '13 at 18:28
  • 1
    Maybe this isn't the answer you're looking for... nobody who has ever done a study like this has wanted to have fewer subjects than they could get. I'd say the answer to your question could just be 'funding' – shigeta Feb 18 '13 at 02:21
  • That's a good point... – Prabhu Feb 18 '13 at 02:50
  • I think it depends on what type of analysis you are going to do, but I know for at least an ANOVA you can calculate the smallest sample size for a given power, but in that case you have to have an idea of how big difference you are looking to find-. – Zewz May 19 '13 at 07:14
  • 3
    About experimental design and is a purely statistical question. – fileunderwater Oct 22 '13 at 11:54
  • 4
    I'm voting to close this question as off-topic because this question is about statistics, not biology. – Remi.b Jun 29 '17 at 15:37
  • @Remi.b Agreed, it is also far too broad for here or Cross Validated. There is an entire tag (although unfortunately populated mostly with other poor questions with no answers): https://stats.stackexchange.com/questions/tagged/sample-size – Bryan Krause Jun 29 '17 at 17:24

1 Answers1

-1

Could be as much as possible – using multistage sampling. But it all depends on how much you are willing to spend, and your time constraint.

Odutolu Timothy
  • 1