I am doing a research on 20 participants where 10 will be from computers background and rest will be from non computing background.

Each participants will attend a lab session and the performance of the participants will be evaluated based on a rubric. Each row of the rubric is evaluated on scale of 1 to 4.

The research question question is: does backgroundof the participant effect to play in learning about xxxx.

Which statistical method should I use to analyze the data?

  • $\begingroup$ Some more information would be helpful in coming to an answer. For eg. How do you define your outcome variable i.e "learning about xxxx". Is it a simple yes/no outcome or an outcome with several categories or an outcome with some kind of a continuous scale? $\endgroup$ – Ariel Apr 9 '12 at 23:00
  • $\begingroup$ I am looking for just yes/no outcome. $\endgroup$ – Dumb_Shock Apr 9 '12 at 23:03
  • $\begingroup$ Do you just have one variable of interest to assess - ie. computing background? $\endgroup$ – Ariel Apr 9 '12 at 23:07
  • $\begingroup$ @ariel : Yes... $\endgroup$ – Dumb_Shock Apr 9 '12 at 23:13
  • $\begingroup$ Well, in this case, I would say the easiest way would be to just compute an odds ratio with a 2X2 table. Your outcome would be Learning: YES/NO and your exposure variable is Background: Yes/No. Quick how to: intmedweb.wfubmc.edu/ebmreviews/odds.html $\endgroup$ – Ariel Apr 9 '12 at 23:20

Based on the information you have provided, you have a binary outcome: Learning Yes/No and a single exposure/explanatory variable (also binary: computing background Yes/no). Calculating an odds ratio could be one quick way to see if the computing background really affects the distribution of your participants in the Learning Yes/No categories. It would give you an answer to this question: What are the odds of learning, given a computing background?

Since this is just a bivariate analysis, I would be very cautious in interpreting the odds since it is possible that you have other hidden variables (that you are not aware of, or are not measuring) that may affect learning in your experiment. If you do have other variables and want to measure their effect as well, I would recommend increasing your sample population size (a rule of thumb is 10-15 more participants per success/failure in learning - whichever is lower) and possibly doing a logistic regression. This will also give you an odds ratio but you can do a multivariable analysis.

  • $\begingroup$ Very nice answer Ariel. i will give you a +1 tomorrow when I have available votes (3 hours from now). $\endgroup$ – Michael R. Chernick Jun 19 '12 at 21:08

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.