Finding the Best Dog Treat with Statistics
6 hours ago
- #statistical-ranking
- #dog-treat-experiment
- #bradley-terry-model
- Bebop the Greyhound loves running, following his owner, and treats. The owner uses the Bradley-Terry model to determine Bebop's favorite treat through pairwise comparisons.
- The Bradley-Terry model assigns strength scores to each treat based on head-to-head matchups, similar to Elo ratings in chess, enabling ranking from comparisons.
- An experiment involves presenting Bebop with two treats at a time, using a 'choice' command, and recording preferences. Treats include MON2SUN duck/chicken rawhide, Greenies, Pork Chomps, and Pur Luv Chicken.
- Results show a right-side bias in Bebop's choices, possibly due to environmental factors. Pur Luv Chicken (Treat E) emerges as the top contender, with MON2SUN duck (Treat A) as a close challenger.
- Bootstrap analysis indicates Treat E wins in 63% of samples, Treat A in 33%, and others rarely. More trials between E and A are needed to settle the ranking.