I have the following experimental design: my treatment is composed out of three factors (factor1, factor2, factor3 ) each having two levels. So far I have eight groups and I assigned subjects randomly to these groups.

To raise the power of the study, I thought it was a good idea to test each subject twice. After testing the first treatment, there is a second treatment to the same person, this second treatment is composed out of all the opposite factor levels.

I don't know how exactly this design is called, looks like a fractional crossover design (or some kind of factional survey experiment) to me.

The response variable is continuous, so I am going for an ANOVA (or should I give the nlme package a try?). Now I am struggling with the right way to do it regarding the following four points.

  1. Does this model formula fit to the design?

    fit = aov(response ~ factor1 * factor2 * factor3 + Error(subject / (factor1 * factor2 * factor3))
  2. And what is the difference to the following formula?

    fit = aov(response ~ (factor1 * factor2 * factor3) + Error(subject))
  3. What if I am only interested in the following effects?

    • Main effect of factor1
    • Main effect of factor2
    • Interaction of factor1 x factor3
    • Interaction of factor2 x factor3
  4. And how do I interpret the output given the above formulas?


Your Answer

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

Browse other questions tagged or ask your own question.