Duplicate subjects in multi-factorial ANOVA

#1
Hello, I have a study design which requires a multi-factorial ANOVA. Basically, I have 3 groups: Patients and two control groups. Each group performed a lower-limb functional test, starting with each leg, and with two different heights (sorry I can't elaborate more). I originally performed four comparisons, for each combination: Right-low. Right-high. Left-low. Left-high. However, due to multiple comparisons, post-hoc corrections ravage my p-values. Plus, I figured a more elegant way may exist. Enter multifatorial ANOVA. I tried to create a model where I have group, leg and height as fixed factors. It works fine, but that means that I have 4 times as many degrees of freedom as participants. This does not feel right. Am I even allowed to do that?
 

Karabiner

TS Contributor
#2
Looks like a "mixed" ANOVA could be performed,
i.e. repeated-measures ANOVA (2 within-subject
factors: leg and height) with an additional grouping
factor (3 Levels).

With Kind regards

Karabiner
 
#3
Looks like a "mixed" ANOVA could be performed,
i.e. repeated-measures ANOVA (2 within-subject
factors: leg and height) with an additional grouping
factor (3 Levels).

With Kind regards

Karabiner
Thanks, but I'm not sure this is what I need. You see, a repeated measures design would test the differences between the different levels of the within group factors. That way, if the patients would have a larger difference between heights or legs than the other groups, the test would be significant. However that is of no interest to me. I am only interested in the overall between-group differences.
Now, if for example all groups have equal low and high results, but the values are different between groups, that won't be discover in the analysis.
 

Karabiner

TS Contributor
#4
You have a repeated-measures design (if I understood your description correctely).
Therefore you have to take into account that the same subjects were measured on
several occasions. Within the proposed mixed ANOVA you can assess whether there
is an overall effect of "group" (across all conditions), and/or whether this effect is the
same across all conditions. It is logically the same as your proposed 3-factorial
ANOVA, except that it takes into account that 2 factors (leg and height) are now
within-subjects rather than between-groups factors.

With kind regards

Karabiner