A1417
Title: Model fitting using partially ranked data
Authors: Mayer Alvo - University of Ottawa (Canada) [presenting]
Xiuwen Duan - University of Hong Kong (China)
Abstract: The importance of models for complete ranking data is well-established in the literature. Partial rankings, on the other hand, naturally arise when the set of objects to be ranked is relatively large. Partial rankings give rise to classes of compatible order, preserving complete rankings. An exponential model is defined for complete rankings, and it is calibrated on the basis of a random sample of partial rankings data. It appeals to the EM algorithm. The approach is illustrated in some simulations and in real data.
