An AB/BA cross-over design randomly assigns half the participants to receive 
two treatments in the order A followed by B while the other half receive in 
the order B followed by A. These data are from an unpublished study of 
diuretics for the treatment of mild to moderate heart failure. After screening,
participants were asked to not take diuretics for a period of at least one 
day and not more than one week until the trial began (pre-trial washout 
period). 

Baseline observations of diastolic blood pressure (taken as the average of three
measures) and oedema (measured by the sum of the left and right ankle diameters)
were taken at the end of this period, and then the treatment order was randomly
assigned. Each of the two treatment periods lasted for five days with no washout
period in between. DBP and oedema were again measured at the end of each 
treatment period.

101 patients were screened for the study, but 7 withdrew prior to randomization.
Only 2 of the remaining 94 failed to complete both treatment periods. 

These data were presented in Brown H and Prescott R (1999). Applied mixed 
models in medicine. Chichester, England: John Wiley & Sons, Inc. They have 
randomly deleted roughly one in five of the second period observations for 
illustration of how fixed effects and random effects modeling can give 
different results in such situations.