I have a set of data from three groups across 3 time points (2 groups are subjected to an intervention and the third acts as control). I want to evaluate intra and inter-group differences between these time points for a given variable. However, at baseline evaluation, one group data is significantly lower compared to the others. I thought maybe I could analyze data from a percentage standpoint, say first assessment is 0% and afterwards intragroup difference is 15% or whatever. I believe this is possible for intragroup analysis but I´m not sure if it is possible for inter-group analysis.
On the other hand, I could just analyze raw data instead of transforming it but this may lead to some issues that I´ll probably have to adress in the discussion.
Any idea on what can I do?
Thanks in advance!
You could use a random intercept term by subject (-> "mixed model", "hierarchical model").
Hello Marcos Quintana Cepedal. What is your dependent variable? If I understand you, it is some raw variable that is not a percentage, but because the baseline level is substantially lower in one group, you wonder if you should be analyzing percent differences (from baseline) rather than raw differences. Is that right? (I remember seeing this same concern in studies analyzing response time data when some groups--e.g., children--responded more slowly overall than another group.) If that is your concern, one approach is to analyze the log-transformed DV. Here is a UCLA page with some relevant info:
- https://stats.oarc.ucla.edu/other/mult-pkg/faq/general/faqhow-do-i-interpret-a-regression-model-when-some-variables-are-log-transformed/
HTH.
There are other choices in the same verin as Bruce suggested. I attached one I found useful. Best wishes David Booth
- Badges
- Science method