I have conducted a randomized controlled trial that compared 2 intervention groups. The outcome measures are the number of observed medications taken (continuous variable) and the proportions of respondents with high medication adherence (categorical variable - high vs low adherence). Several independent variables could interact with the outcome, including sociodemographic, income, and social risk factors etc., that need to be adjusted to find whether any of the intervention groups effectively improve medication adherence. What statistical analysis is suitable for this kind of data? Thanks in advance
For continuous variable:
- Use linear regression with observed medication as dependent variable and intervention group as independent variable.
For categorical variable:
- Use logistic regression with high/low adherence as dependent variable and intervention group as independent variable
I would argue that if your randomization went well, there is actually no need to adjust for covariates across groups :-)
You should be able to answer your research question by running two tests, e.g. Mann-Whitney for the number of medications taken (check whether a difference in distributions is what you're after though), and Chi-Square for the second, as this is just a 2x2 table if I understand correctly.
You could also use a regression model in the first case, which BTW sounds more like count data to me, but you would have to check whether the distributional assumptions hold in your data.
- Badges
- Science topic
- Similar topics
- Heat