The advantage of repeated measures

The difference between ordinary and repeated measures ANOVA, is similar to the difference between unpaired and paired t tests. See the advantages of pairing or matching. Since each participant or experiment acts as its own control, repeated measures design can do a better job of separating signal from noise, so this design usually has more power. Some participants may have larger measurements at all time points, and others may have lower measurements at all time points. Repeated measures ANOVA focuses on how much the Y value changes between treatments.

Repeated measures or randomized block?

The term repeated measures is used when you give treatments repeatedly to each animal or participant.

The term randomized block is used when you randomly assign treatments within each cluster (block) of matched subjects.

Imagine that you compare three different treatments. In a repeated measures design, you'd recruit say 10 participants (or use ten animals) and measure each of the participants (animals) after each of the treatments. With a randomized block design, you'd recruit ten clusters of four partipants each, matched for age, gender, postal code, etc. (or ten sets of four animals, with the four treated at the same time in adjacent cages...). Another example would be a laboratory experiment is run several times, each time with several treatments (or a control and several treatments) handled in parallel.

Any time the participants/animals/experiments are clustered in some way that can make their responses similar, other than the treatments you’re comparing, you need to treat those clusters as randomized blocks. Examples include animals from the same litter, blood samples from the same draw, or mice tested in batches on the same day.

ANOVA works identically for repeated-measures and randomized block experiments, and Prism always uses the term repeated-measures.

Counting factors

One-, two- or three-way ANOVA? Two possible points of confusion:

•A design with three or more measurements of different treatments on the same participant is called repeated measures one-way ANOVA in Prism, because there really is only one factor, treatment as denoted by the data set columns. But you could argue there is a second factor too, subject, because each row represents a different subject (or block). In fact, you'll get the same results if you analyze with two-way ANOVA (without replicates) and one-way repeated measures ANOVA.

•Say you give some participants one treatment, others a different treatment,  and others a third treatment. You then  measure an outcome for each participant at three time points (before, during and after the. There are two factors: treatment and time. So you need two-way ANOVA, with repeated measures in one factor. Some people mistakenly ignore time and think this is a one-way ANOVA problem, forgetting that time is a factor.