2
$\begingroup$

I could need some help on how to choose the correct way to analyze the data I have collected.

I conducted an experiment with two groups (control condition vs. experimental condition). The continuous dependent variable was tested twice after the intervention for each subject. That is, I have one between subjects-factor (condition) and one within subjects-factor (test time). Additionally, I have one continuous control variable that ideally I would like to include in my model, but this is not a priority.

Basically, I would like to test the following hypothesis:

  • Is there a difference between the two conditions at test time 1?
  • Is the change between test time 1 and test time 2 different for the two conditions (interaction condition/test time)?

These are the ideas I have come up with:

A hierarchical linear model/multilevel model

...with time as a within subjects factor, condition as a between subjects factor and an interaction of both to analyze the change between test times 1 and 2.

Advantage: I can include my control variable and all hypothesis can be tested in one model.

Disadvantage: Is it correct? Plus, my supervisor/reviewer would prefer a more simple approach.

A repeated measures ANOVA

... again with time as a within subjects factor, condition as a between subjects factor and an interaction of both to analyze the change between test times 1 and 2.

Advantage: Similar to multilevel approach, I could include my control variable.

Disadvantage: I'm not sure if it is correct. Additionally, it may still be too complicated. And I would need to perform post-hoc analyses.

Two t-tests

One t-test for the difference between the two conditions at test time 1. Then I would calculate the difference between test time 1 and test time 2 and in a second t-test would test if this difference differs between the two conditions.

Advantage: "Simple". If I put the variables in an ANCOVA I could additionally control for my continuous variable.

Disadvantage: Wrong? Okay? I'm not sure about it.

Summary

I have to meet two demands: I want to analyze the data correctly and I have to do it as simple as possible. Thus, if it would be okay/correct to analyze the data with the two t-tests approach I would probably go for it, as my supervisor/reviewer would be happy with it. However, if it is not correct, I would be okay with one of the "more complicated" approaches, if I were at least sure that they are okay.

$\endgroup$
1
$\begingroup$

Today, I'm pretty sure that the correct way to analyze this kind of data is a multilevel regression model (https://neuropsychology.github.io/psycho.R/2018/05/01/repeated_measure_anovas.html). I chose a random intercept model where the intercept varies per subject (because each subject was observed twice) and included the treatment, time and control variable effects as fixed effects.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.