We have 50 sample convinently select, outcome measured at baseline (M1), then exposed to intervention A for 6 weeks and measured (M2), they were allowed to undergo 1 week of wash out effect and a new baseline was taken (M3), then exposed to intervention B for 6 weeks and final measurement was taken (M4).
1. Can I call this a one group crossover trial?
2. Can I use a 2 ways repeated measure ANOVA to analyze for (group X time) differences?
3. In case of non paramateric can I run an ordinal or logistic regression?
Dear Bruce Weaver ,
why do you think that this can be calleda "cross-over"? There is nothing crossed over. It's asimple sequential design: first A, then B.
If the wash-out is not effective, it is very difficut to interpret the data. But even if the wash-out works well, what cannot be solved here is the problem that the time itself may have an effect on the response or -more serious: the effects of A and B. The experiment runs over several months, and the climatic changes during this time and exposures to different viruses coming up at different times of the year, and a change in diet (if seasonal food is used) can easily have a relevant impact.
Taking these problems aside, there are 4 measurements per subject (M1...M4). The effect of A is given by the diffreence M2-M1 (for each subject), and that of B is given by M4-M3. The Difference between A and B is the differenceof these differences, (M4-M3) - (M2-M1), again per subject. These differences can be tested with a one-sample t-test or a sign test (depending on the kind of data).
Hello Ogochukwu Kelechi Onyeso. Your question reminds me of something Robert P. Abelson said in his book, Statistics as Principled Argument--see the attached PDF. Now to your specific questions.
1. Yes, you could call it that.
2. If you call it a one-group crossover trial, then "group" cannot be a factor in your two-way RM ANOVA. Call it "Treatment" instead (A vs B). Referring back to Abelson, the criticism that is likely to arise is that you cannot know if the order in which A and B are administered has any effect on the results. Reviewers will likely ask why you did not randomly allocate the 50 Ss to two order groups AB and BA. Then you could conduct a 2x2x2 (Order x Treatment x Time) ANOVA and test whether Order interacts with any of the within-Ss effects. (If you have not actually collected the data yet, I suggest you do this.)
3. It is not clear what you are asking here. What is the outcome variable? If it makes sense to use means and SDs to describe it, ANOVA ought to be fine for inference.
HTH.
Dear Bruce Weaver ,
why do you think that this can be calleda "cross-over"? There is nothing crossed over. It's asimple sequential design: first A, then B.
If the wash-out is not effective, it is very difficut to interpret the data. But even if the wash-out works well, what cannot be solved here is the problem that the time itself may have an effect on the response or -more serious: the effects of A and B. The experiment runs over several months, and the climatic changes during this time and exposures to different viruses coming up at different times of the year, and a change in diet (if seasonal food is used) can easily have a relevant impact.
Taking these problems aside, there are 4 measurements per subject (M1...M4). The effect of A is given by the diffreence M2-M1 (for each subject), and that of B is given by M4-M3. The Difference between A and B is the differenceof these differences, (M4-M3) - (M2-M1), again per subject. These differences can be tested with a one-sample t-test or a sign test (depending on the kind of data).
That's a fair point, Jochen Wilhelm. If you make a diagram of the design, there is no cross-over. So you're right: it should just be called sequential. Thanks for correcting me! ;-)
- Badges
- Science topic
- Similar topics
- Methodology
- Research Methodology