I am conducting a longitudinal study where three groups are measured over 4 different time points (T1, T2, T3 and T4). I am mainly interested in what happens within groups.

But to see if the percentage change between every time point is different between groups, I was reading about the statistical meaning of percentage change. And one article I read concluded that percentage change, although often reported, should not be used in statistical analyses (source:Article The use of percentage change from baseline as an outcome in ...

Roughly summarized: If you do want to report percentage change. You should first do an ANCOVA between groups with post values as dependent variable and pre values as covariate. Afterwise you can report percentage change by dividing the (mean post-value of the group adjusted for the pre-values - mean pre-value of the group)/(mean pre-value of the group)*100. In that way you can compare the percentage change between groups.

This is a great solution, but only holds for pre-post designs. My design has 4 repeated measures. Should I do three ANCOVA's then (comparing T2 with T1 as covariate, comparing T3 with T2 as covariate and comparing T4 with T3 as covariate)?

Or is there another way? I am thinking about mixed linear modeling for example an adding the pre value as a covariate. This raises another question: do I only use the first baseline value as a covariate (a non changing one), or do I use a changing covariate (value T1 as covariate when comparing T2 between groups, value T2 when comparing T3 between groups, value T3 when comparing T4 between groups).

Any advice is welcome.