Hello,
I am hoping to get some suggestions about how to best organize some of my variables. Some background, my study groups subjects by the sports they play (6 sports total) and the outcome variables are on a Likert scale. I plan on doing a Kruskal Wallis test followed by a post hoc Dunn's test to test for differences between sports.
The issue I have is that there is overlap in my groups (some subjects play more than one sport).
What would be the best way to go about organizing my data to minimize confound?
Thank you!
This may be using a sledge hammer to crack a nut but you have a non-hierarchical (multiple membership or classified) structure and there are quantitative methods to deal appropriately with that:
see structures and classification in
http://www.bristol.ac.uk/cmm/software/mlwin/mlwin-resources.html#mlmbkgrd
and more detail
http://www.bristol.ac.uk/cmm/software/mlwin/mlwin-resources.html#nonhier
Would it be appropriate to duplicate the subjects with overlapping groups by the number of sports they play to obtain separate groups for this part of the analysis only (not for demographics or overall sample size)?
For example, subject 100 plays baseball and football, so create 2 separate rows and in my "sport" column, indicate football in one row and baseball in the other row?
Akash R. Patel, I think you need to provide more information. For outcome variable Y, is subject 100 asked to give just one response, or is he (or she) asked to give one response with respect to baseball and another response with respect to football? Thanks for clarifying.
Bruce Weaver Thanks for your response. The subjects were asked to give only one response to different questions (all on a Likert scale) about their behaviors while playing "sports in general", and we asked them to report which sports they played.
Thanks for clarifying, Akash R. Patel. You definitely do not want to create additional rows of data for subjects who played more than one sport. This would create what Snijders & Bosker (2012, p. 16) describe as "the miraculous multiplication of the number of units".
I think I would start by creating indicator variables for each of the 6 sports (1=plays, 0=does not play) and cross-tabulating them to get an idea about which combinations of sports are most prevalent.
What is your sample size?
- Badges
- Science topic