Cross-over experiment

  • 15 participants
  • 2 treatments: active, sedentary

Question:

  • How does treatment affect the amount of calories consumed?

Design

Two weeks of each

  • control/washout
  • randomly assigned treatment
  • washout
  • other treatment

Measurements

Measurements made at 3 times:

  • before 1st treatment
  • after 1st treatment
  • after 2nd treatment

At these times, the participant is given spaghetti for breakfast. The participant fills out forms and then is given pizza for lunch. The amount of pizza consumed is the main measurement of interest.

Proposed Model

Response:

  • amount of pizza consumed

Explanatory variables:

  • treatment
  • period (when the observation was taken)
  • sequence (the order the treatments were given)

Random effect:

  • person

Question

Should we include the baseline measurement as an explanatory variable in the model?

Feedback

  • We believe the measurement could be improved.
  • This seems underpowered.
  • The goal of the baseline measurement is to control for differences between people, but the person random effect already controls for differences. Thus it seems there is no reason to additionally include the explanatory variable.