I have two groups of samples: healthy and diseased groups. Each had an X treatment. So, I have a healthy baseline and healthy X treatment. The same for the disease group. Which is the best way to analyze them? If I use one-way ANOVA, it only measures the mean. But it doesn't take into account that they are paired measurements. The healthy group did change, but its mean is lower; therefore, it is not assumed to be significant. If I do a paired T-test, I get how each group responded to the treatment but cannot compare healthy vs disease. Hope you can help me.
3
- $\begingroup$ It sounds like you want to compare healthy and disesed individuals, before and after the treatment? This sounds like a 2 x 2 repeated measures ANOVA. If you are using R this is easiest using nlme or lmer with a model like lmer(response ~health*treatment + (1|ID)) $\endgroup$N Brouwer– N Brouwer2024-04-06 21:54:16 +00:00Commented Apr 6, 2024 at 21:54
- $\begingroup$ It depends on what you are trying to prove. To show that traetment X improved the Diseased group more than the Healthy group (if they are Healthy, treatment X should not have a big effect), you could compute the "Before-After" measures for both groups (Healthy and Diseased), and then use a 2-sample t-test (Welch version) to compare the mean change. If you are trying to prove that treatment X was effective, just run a paired t-test on the Diseased (but then you should have used a placebo group, not a healthy group). ...cont... $\endgroup$jginestet– jginestet2024-04-07 00:58:52 +00:00Commented Apr 7, 2024 at 0:58
- $\begingroup$ ...cont... If you are trying to prove that the Diseased are "cured" (as good as Healthy), run a 2-sample t comparing Diseased after to Healthy before. Honestly, it is not clear from the question what the purpose of the study/data is. Can you clarify? $\endgroup$jginestet– jginestet2024-04-07 00:59:02 +00:00Commented Apr 7, 2024 at 0:59
Add a comment |
1 Answer
$\begingroup$ $\endgroup$
You can use at least
repeated-measures mixed ANOVA with healthy vs. disease as between-subject factor and time as within-subject factor,
Analysis of Covariance (ANCOVA) with baseline scores as a covariate, healthy vs. disease as between-subject factor and post-treatment scores as the dependent,
(here is a very clear presentation regarding the differences between 1 and 2 when dealing with pre-post data; note that this deals with the more typical situation where the treatment (control vs. treatment) is the between-subject factor and the groups are initially similar whereas you have initially two groups and the treatment is the same. However, the analyses suggested work for your data too).
- multilevel regression with time*health group interaction and a random intercept of participant as suggested by N Brouwer in a comment. In your situation this is pretty much the same thing as repeated-measures ANOVA (1), but multilevel regression handles missing data better than ANOVA.
