Testing differences between groups over time


I am writing a research proposal that looks at a change in scores over time between three groups. Here is the basic set-up:

Three randomized groups (2 intervention (central and peripheral), 1 control)

Pre-test baseline measures (trait measures), pre-intervention state measures, post-intervention state and trait measures, and 3 month-follow-up (trait) measures

Women will complete baseline measures online. Then they will come into the lab, do pre-test state measures, do the writing assignment (intervention), complete the post-intervention state and trait measures, and then complete the baseline measures 3 months later.

Here are my hypotheses:

Immediately following the intervention:
1) Women in the central and peripheral conditions will experience increased state body appearance and body functionality satisfaction relative to the control condition, with women in the peripheral condition displaying the largest effects.
2) Post-test body appreciation scores will be higher in women in the central and peripheral conditions relative to the control condition.

It is hypothesized at follow-up that:
3) Body appreciation and body satisfaction scores will remain stable at follow-up for women in the central condition but will attenuate in women in the peripheral condition (as research demonstrates that changes via the central route are longer lasting whereas effects via the peripheral route are more vulnerable to change; Baumeister & Finkel, 2010).

So, for the first hypothesis, I proposed doing a mixed-method repeated measures ANOVA with 3 levels as the between-subjects factor (condition) and 2 levels of the within-subjects factor (pre-intervention and post-intervention). The scores on the state measure will be the dependent variable. I know a lot of people use ANCOVA here but, I have heard that ANCOVA should only be used for failure of random assignment. Thoughts?

I am not sure the best way to test that one of three conditions will show the strongest effect-do I do a post-hoc Tukey's or a planned comparison?

For the second hypothesis, I planned to do a simple one-way ANOVA with a planned comparison where I check that the weighted mean of both intervention conditions is significantly different than the control condition.

The third hypothesis is where I am stuck. I am proposing two mixed-methods repeated measures ANOVA with condition as the between-subjects factor, baseline, post-intervention, and follow-up as the within subjects factor (3 levels), and scores on the trait measures as the dependent variables. Right now I'm proposing to visually inspect the graphs if there is a significant interaction. Is there any other way to test that hypothesis?

Am I totally off base? I usually live in regression land so ANOVAs and such are a bit of uncharted territory for me. Any help is appreciated!
ANCOVA is not only used when there is an assignment failure. It is used like an ANOVA, only that you wish to adjust for some meaningful covariates that you cannot manipulate.

For your design, a repeated measures ANOVA with planned comparisons is appropriate. Remember to adjust for multiple comparisons.

Lastly, Regression, ANOVA, ANCOVA, MANOVA, etc all belong to the same family of statistical methods called, "General Linear Models". Aka if you include the same variable into your regression analysis, the exact same result would appear in any other of the abovementioned modeling techniques.