Dear STATA list user,
I am currently working with a 2-stage crossover design testing a new innovation (medical equipment), which requires repeated measurements of oxygen saturation at 0 sec, 30 secs, and 60 secs). The outcome variable is oxygen saturation or 'sat'. The treatment variable is 'intervention'. The 3-measurement points will be treated as continuous variable, assuming linear relationship, referred to as variable ''measure'. For crossover design, the sequence or the order of the given innovation is mandatory, so the order of treatment given will be stored in variable 'order'.
number variable: patient id (cluster marker)

Treatment effect for the innovation:
Code:
xtmixed sat i.intervention##c.measure || number:measure,
Period effect (Crossover design)
Code:
xtmixed sat i.order || number:measure,
Sequence effect : interaction of treatment effect and order of treatment
Code:
xtmixed sat i.intervention##c.measure##i.order || number:measure,
(as treatment effect in this scenario is "i.intervention##c.measure")

Finally, I incorporate treatment effect, period effect and sequence effect within the same equation. The command is as follow:
Code:
xtmixed sat i.intervention##c.measure i.order i.intervention##c.measure##i.order || number:measure,
the yielded Stata result was the same as:
Code:
xtmixed sat i.intervention##c.measure##i.order || number:measure,
Does this look like a correct analysis?
Should I use three-way interaction to analyze a crossover design with repeated measurements?
Does the term "i.intervention#i.order" have any meaning? because 'i.intervention' is not the slope?

Thank you very much in advance