Outliers in Probability Models

J. S. Cramer, J. S. Cramer
1985 unpublished
I propose a likelihood ratio criterion for identifying outliers among individual observations to which a probability model has been fitted. Outliers are found to occur especially among the rare sample categories, and further analysis suggests that the fit of a probability model is systematically biased in favour of the predominant categories. I intend to pursue the analysis with drawing further conclusions from Figure 1, and this may possibly lead to a proposal for (re)weighting observations in
more » ... ing observations in the estimation of a probability model. The analysis is throughout illustrated by the application of a multinomial logit model to four automobile ownership modes of some 3 000 households from a budget survey.
doi:10.22004/ag.econ.302173 fatcat:p2yoaohdanc7bewmkzy6iiqhqq