I have a non-random sample of undergraduate students (n-230). The aim is to conduct a mediation-moderation approach. I would like to know my options for overcoming selection bias beyond including control variables. I have read about propensity scores but I do not have a case-control design. Is bootstrapping an option? Others?
You have about got it unless want to take a random sample
David Booth
Hello David,
There is no error-free solution (as implied by David Eugene Booth 's response).
About the best you could do would be to compare your sample on an array of attributes with those of the target population, assuming such information either exists or could be obtained. Ideally, your sample would not vary in any or all of those attributes from the target population.
However, while this approach may give you confidence that your sample is not wildly different from the target population, at least with respect to the attributes compared, it does not offer a blanket guarantee of representativeness.
Good luck with your work.
David Eugene Booth and David Morse thanks for the guidance and insight!
- Badges
- Science topic
- Similar topics
- Biological Science
- Histology
- Sectioning