I am doing the evaluation for a program with 200 members (they enrolled in the program at different times in the past year and the program doesn't end). Only 78 completed the pre test survey. Now I have to look for a control group and resources are scant... What is the ideal number of a sample size for the control group? What is the optimal number? I also know it is a big limitation that treatment and control group didn't take the pre test at the same time, how big is that limitation? Thanks! M.

  • $\begingroup$ It is rather hard to say in the abstract. What kind of data do you have? Eg, do you think of the response as continuous? How big of a difference between the treatment & control groups do you want to be able to detect? $\endgroup$ – gung - Reinstate Monica Jun 29 '15 at 14:51
  • $\begingroup$ Yes, my main health outcomes are BMI, blood pressure, and glucose level. $\endgroup$ – Mariana Jun 29 '15 at 15:13

In epidemiological research an optimal ratio is of 4 controls to 1 case (assuming it's an unmatched design). After that, power increases but at a decreasing rate as the ratio of controls/cases increases.

enter image description here

from Woodward, M. Epidemiology Study Design and Analysis. Boca Raton: Chapman and Hall:, 1999, p.265

| cite | improve this answer | |
  • $\begingroup$ Thanks Marijana. Although this is not an epidemiological research. This is a health program, focused on wellness and prevention... $\endgroup$ – Mariana Jun 29 '15 at 15:14
  • $\begingroup$ hi Mariana, I think the same methodological issues apply to public health intervention. What effect size do you wish to achieve with your program? e.g. odds ratio of 1.5, or 2? What is the prevalence of the risk factors. With this information, you can use standard online sample size calculators, like sampsize.sourceforge.net/iface/s3.html#ccp $\endgroup$ – Marijana Vujkovic Jun 29 '15 at 15:49

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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