Cross-over design

20 subjects were randomly assigned to one of two sequences of music:

  1. rest -> relaxing -> activating
  2. rest -> activating -> relaxing

During each music period (rest, relaxing, and activating) for each person, 20 impulse measurements were taken.

The client appears interested in

  • differences in means between relaxing/activating and rest
  • differences in coefficient of variation between relaxing/activating and rest

The client is also interested in the effect of years of education.

The client is fitting random effects models in SPSS.


Plot the data

With this little data, we can easily plot all of the data with time on the x-axis and impulse measurements (probably logged) on the y-axis and individuals faceted by sequence.

Summarize each individual-treatment combination

We have the following suggestions for responses all of which are computed on each of the 20 measurements in an individual-treatment combination:

  • mean of the logged measurements
  • coefficient of variability
  • standard deviation of the logged measurements

We prefer the standard deviation of logged measurements to the coefficient of variability because interpretation of the resulting analysis will be easier.

For each of the desired responses, analyze using a linear model with random effect for individual and fixed effects for treatment and sequence.