Fitting loglinear Bradley-Terry models (LLBT) for paired comparisons using the R package prefmod

This paper aims at introducing the R package prefmod (Hatzinger, 2009) which allows the user to fit various models to paired comparison data. These models give estimated overall rankings of objects or items where each subject (respondent/judge) makes one or more comparisons between pairs of objects (items). The focus is on the loglinear Bradley-Terry (LLBT) model, the loglinear formulation of the Bradley-Terry(-Luce) model, both assuming independence between comparisons. Five types of data structures are covered: (i) simple paired comparisons, (ii) paired comparisons including an undecided category, (iii) categorical subject covariates (for estimating different overall rankings for different subject groups), (iv) object covariates for reparameterizing objects, and (v) order (position) effects. Additionally, the discussion briefly addresses other response formats such as ratings and (partial) rankings.