Hello all, I am working on a project that is comparing Likert scale data "before" and "during quarantine." My team is interested in comparing these responses across geographical regions to assess how different regions of the US may have impacted participants in their quarantine/lock-down response. In short, I'm wondering if there is a way to compare ordinal data across categorical data across two time points (continuous data).
For simplicity, I am leaning towards using Wilcoxon signed rank test to compare Likert scale "before" and "during quarantine." Friedman's or Cochran's wouldn't be appropriate because we did not collect more than 2 time points. From there, I could run another analysis on "during quarantine" vs geographical region. Ideally, I'd like to compare all three variables at once and run post-hoc tests to assess magnitude of change.
Any advice will be appreciated.
If you want to treat your dependent variable as ordinal, the best approach with your setup is to use ordinal regression. Depending on the software you are using, this should allow you to account both for the geographical region, the repeated measures aspect of before and during, and the interaction of these, all in one model... Note that if you have a true "scale" that is composed of multiple questions, you may want to treat your dependent variable as a continuous variable.
The only thing I disagree with Salvatore on is ordinal is ordinal and never ever normaly distributed because the dependant variable is not continuous ever check the definition of continuous variable. Best wishes, David Booth
Although I'm a bit confused at your explanation. More details is required because it sounds like a two way ANOVA can suit your problem if the dependent variable is continuous. Besides is the Likert scale data collected before and during the quarantine your dependent variable? If yes, then run a one-way Anova comparing the regional differences before quarantine. Once you are done, run another separate one-way Anova
comparing the regional differences on the measured attribute during quarantine.
Another alternative is to use one-way Analysis of Covariance. Use the regions as the independent variable. Then the data of the measured attribute before quarantine can be used as the covariate. Data of the measured attribute during quarantine can be used as the dependent variable. This should handle your problem.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size