0
$\begingroup$

My experimental design is the following.

I have 4 experimental groups, with each drug alone or combination (Control, DrugA, DrugB and DrugA+B) and these treatments were administered after surgery. I run a test before surgery (baseline) and again (on the same subject) after surgery/treatment.

Is two-way repeated-measures ANOVA the appropriate test in this case?

$\endgroup$

1 Answer 1

2
$\begingroup$

From what you describe, it sounds like a two-way repeated measures ANOVA (with interaction) is the right choice. Although, if "Control" is something other than "Neither Drug A nor Drug B", then I would probably run this as a one-way repeated measures design with four treatment levels. I think this can be seen nicely if you represent your factors in as a table such as:

                       Drug A
                  No             Yes
            ----------------------------------
Drug B  No  |  Control   |      Drug A       |
            ----------------------------------
       Yes  |   Drug B   |  Drug A + Drug B  |
            ----------------------------------

So, if your Control is something else, then this design may not be proper for your experiment.

$\endgroup$
2
  • $\begingroup$ Control is indeed "Neither Drug A nor Drug B". I ran the two-way repeated measures ANOVA on graphpad (only program available in all computers in the lab) and DrugA+B is significant compared with Control. @Jelsema Is there any difference between running this on graphpad and SPSS? Can this 2x2 treatment scheme showing my experimental groups be more accurately shown with SPSS? In graphpad I can just put 4 different treatments and I feel like if they were control plus 3 different drugs the result would have been just the same. $\endgroup$
    – Mariana M
    Oct 1, 2015 at 8:50
  • $\begingroup$ I have never used GraphPad, and it's been quite a while since I've used SPSS at all, so unfortunately I can't comment on the differences between those programs. That being said, there can be multiple ways to implement models. Treating this as a two-factor design, or as a one-factor design with 4 treatments, should get the same results. The two-factor design might make some follow-up analyses slightly easier to obtain (e.g., if the model is determined to be significant, assessing the individual drugs, and their interaction). $\endgroup$
    – Jelsema
    Oct 1, 2015 at 15:37

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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