two way anova or two way repeated measurments ANOVA


I had a pool of data and I used this data to test 4 models at 4 different starting points. (the data or longitudinal)
--> so this would explain why i'm interested in a two-way ... (models and startingpoints)

The point however is as follow: not all data made it to every group and there is some overlap but not every data in group 1 is present in group 2. So all the has the same origin (my pool of data) but the other way around this doesn't count.

(from my experiences with paired t-tests you compare F-values in this situation and if F values match then you can perform a paired t-test even if a point x is not present in one group)

My questions are as follow:

- Does the same assumption goes for a repeated measurments ANOVA? (that not every datapoint should be in every group)
- If so what's the "F-test" needed to see that the groups can be compared?
- Could using the two-way anova be an option because independency of observations or can i violate this assumptation?

Thanks in advance