This week's speaker at the Gov 3009 workshop is Jasjeet Sekhon from the
Government Department. He will be speaking on "Robust Estimation and
Outlier Detection for Overdispersed Multinomial Models of Count Data."
The full paper can be found at:
http://elections.fas.harvard.edu/election2000/MebaneSekhon.multinom.pdf
Abstract:
We develop a robust estimator---the hyperbolic tangent (tanh)
estimator---for overdispersed multinomial regression models of count
data. The tanh estimator provides accurate estimates and reliable
inferences even when the specified model is not good for an unknown
minority of the data. Seriously ill-fitted counts---outliers---are
identified as part of the estimation. A Monte Carlo sampling
experiment shows that the tanh estimator produces good results at
practical sample sizes even when ten percent of the data are
generated by a significantly different process. Theoretical results
suggest that asymptotically the estimator will produce good results
when up to half of the data are contaminated. The experiment shows
that, with contaminated data, estimation fails using four other
estimators: the nonrobust maximum likelihood estimator, the additive
logistic model and two SUR models. Using the tanh estimator to
analyze data from Florida for the 2000 presidential election matches
well-known features of the election that the other four estimators
fail to capture. In an analysis of data from the 1993 Polish
parliamentary election, the tanh estimator gives sharper inferences
than does a previously proposed heteroscedastic SUR model.
Seminar Information:
The seminar meets at noon in Room 22, Center for Basic Research
in Social Sciences (CBRSS, 34 Kirkland St., this is the yellow building
across the street from William James Hall). Contact information, previous
presentations, and the spring schedule may be found at the course web
site:
http://www.courses.fas.harvard.edu/~gov3009/. Lunch is provided.
To join the gov3009 mailing list, send e-mail to
gov3009-l-request(a)fas.harvard.edu with the following text message:
subscribe
end
Questions? Please contact the workshop coordinator, Liz Stuart, at
stuart(a)stat.harvard.edu