0
$\begingroup$

I've been asked to review a study comparing two groups, in which Group A=43 subjects and Group B=15,000 subjects. Intuitively, I feel that this is not a valid design, but can't find any specific references indicating this. Group A is a disease group, Group B is a healthy comparison group, both sampled from the same national survey. Can anyone provide references or comments about how to handle extreme unbalanced designs such as this.

Improve this question
$\endgroup$
1
  • $\begingroup$ So would you describe this as a (retrospective) case-control study? There will be the inherent limitations of that study design. Are the groups well matched in terms of potential confounders (eg age, sex, socioeconomic status)? $\endgroup$
    – tristan
    Mar 11, 2015 at 19:13

1 Answer 1

Reset to default
1
$\begingroup$

There is nothing inherently wrong with imbalanced groups like this. Certainly, if one can find only a few people with a disease, and has access to values on the same variables for a great many people without the disease, it's better to include all the comparison subjects in one's analysis than remove some of them.

It's true that, given a fixed sample size, most analytic methods benefit from group sizes that are closer to equal. But this was not an experiment—the researchers didn't get to choose which subjects ended up in which group—so they're not to be blamed.

More specific issues may apply depending on the specific analyses conducted and how the sample was collected.

Here is a paper discussing imbalanced data generally:

He, H., & Garcia, E. A. (2009). Learning from imbalanced data. IEEE Transactions on Knowledge and Data Engineering, 21, 1263–1284. doi:10.1109/TKDE.2008.239

Improve this answer
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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