Cross-over behavioral experiment

The purpose of this experiment is to determine preference for light source.


Two light sources:

  • LED
  • CFL

A cage has two rooms connected by an open door. One room has an LED light and the other with CFL. The experiment is run over 4 days with the light source alternating each day. The experiment is repeated 6 times with different birds.


  • Number of bird hours under a particular light
  • Number of birds per cage
  • Amount of feed consumed


When analyzing the experiment first try to calculate a one number summary for each trial-cage combination, e.g.

  • Average (over days) number of bird hours on the LED side
  • Average (over days) number of birds per cage at night
  • Average (over days) difference in feed consumed (LED-CFL)

Then determine whether these are significantly different than a pre-specified value, e.g.

  • 15 bird-light hours
  • 1.5 birds per cage
  • 0 difference

If the signal is big enough, then we will get significant differences here. If not, then we may try to build more complicated models that account for

  • light order
  • day
  • trial (repeat)
  • cage side

But then we need to deal with random effects and correlation which could be negative.


One concern is the antagonism between birds. If there is a bully bird and the other two avoid it, then when you measure the cage, you will believe that one light source is perferred but in fact the opposite is true.


Here are some references concerning whether to keep analyses simple: