HI there,
I am looking to examine whether a categorical variable (e.g., gender) moderates the difference in food intake between two time-points. How can I do this?
Could I compute a change score, and compare the means for that change score between genders?
Or could I regress the score at the second time point on gender and control for the first time point to indicate the impact on the change?
Thanks a lot,
Esther
Think of ideas like analysis of Covariance and Repeated Measure.
Hello Esther,
If you believe that difference scores (e.g., time2 measure - time 1 measure) are error-free, then feel free to compute difference scores and compare them statistically between gender groups (ordinary independent t test could work). However, note that gender may be a proxy for any number of confounding variables which themselves may be the "actual" moderator.
An alternative design, which fares a bit better if measurement error is a possibility, is to use time 1 values as a covariate, and run a simple ancova on the time 2 scores.
Good luck with your work.
Hi David,
Thanks a lot for your reply - my scores are prone to measurement error, so I think this may be a suitable option.
However, what is the difference between using an ANCOVA as you said and a repeated measures approach? (With time as the within-subjects factor, the two time-point scores as the within-subjects variables and gender as the between-subjects factor?) I have tried this and the output gives me a significance value for time*gender.
Thanks again,
Esther
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Regression