I'm new in the field and I would really appreciate if someone could help we with the following statistical problem:
We have a simple repeated measures design with two groups:
Group A: 9 subjects
Group B: 25 subjects
5 measurements (arm swing amplitude) were taken from each subject at a different walking speed levels.
We also have 2 covariates (age and gender). Our main interest is the between group difference in arm swing amplitude.
The problem is the following: all subjects from group A are genetically related and carriers of a certain mutation (PINK1), all control subjects from group B are non-related. As such an rmANCOVA approach does not really work due to the violation of the independency assumption between subjects in group A. Although AN(C)OVA has been largely used in the literature in similar designs, a reviewer asked us to account for independency in the data. Am I correct that this problem needs to be addressed within a multilevel framework (with two levels of clustering)? Is this problem solvable at all given the fact that there are only two groups?
We tried the following approach:
A a new grouping variable (genetic relationship, GR) was defined and subjects ordered accordingly. The labeling resulted in 26 clusters of data, each of the first 25 clusters containing 1 (healthy) subject and the 26th cluster containing the 9 mutation carriers. Model estimation was performed with the software package “nlme” for “R” using the command “lme” and the following model equation:
Amplitude ~ Speed*Group + Age + Gender, dat speed levels, an interaction term between subgroups and speed levels, age and gender as fixed effects and “subjects nested within clusters of GD” as random effect.
Is this ok are there other solutions?