Kenny Shirley
Postdoctoral Fellow
Earth Institute & Applied Statistics Center, Columbia University
My CV (December
2008).
_______________________________________________________________________________________________________________________
ABOUT ME
I am a second-year postdoc at Columbia University in the Earth Institute. My advisor is Andrew Gelman, who is a professor with a joint appointment in the Statistics and Political Science departments, and is the director of the newly-created Applied Statistics Center here at Columbia. Here are some links to places I work:
Earth Institute, Columbia University
Applied Statistics Center, Columbia University
Statistics Department, Columbia University
Statistics Department, Wharton
School, University of Pennsylvania (where I went to graduate school)
_______________________________________________________________________________________________________________________
RESEARCH
(1) Hidden Markov models for longitudinal data.
This work was the focus of my dissertation in statistics at the University of Pennsylvania (2007, advised by Dylan Small). We fit a hierarchical Bayes hidden Markov model to data from a clinical trial of a drug to treat alcoholism. The data were daily time series of alcohol consumption for a large group of individuals, where the hidden states in the model correspond to different “drinking behaviors”, such as binge drinking, steady, moderate drinking, etc. Here is a draft of our first paper, which we are currently revising and expanding.
Hidden Markov Models for Alcoholism Treatment Trial Data (Shirley, Small, Lynch, Maisto and Oslin).
Presentations:
May 2008. Statistics in Psychiatry Symposium, New York, NY. “Bayesian hidden Markov models for alcoholism treatment trial data” (slides pdf)
Previous presentations: JSM, August 2007, Salt Lake City, UT; ENAR March 2007, Atlanta, GA; JSM August, 2006, Seattle, WA.
(2) Models for rainfall and applications to index insurance.
This is a research project I’m working on with Dan Osgood and Andy Robertson at IRI (Columbia University). We are fitting statistical models to daily time series of precipitation in order to accurately design and price index insurance contracts for small farmers in developing countries. Index insurance is a type of insurance that provides payouts based on an index, such as the amount of rainfall at a given location, rather than the actual risk it insures against (crop failure). The additional complication of having the payouts depend on an index is balanced by the fact that index insurance generally avoids moral hazard and adverse selection, two major problems with traditional insurance.
Presentations:
October 2008. Technical Issues in Index Insurance, New York, NY. “Rainfall modeling and simulation” (slides pdf)
Previous presentations: ISBA 2008, Hamilton Island, Australia (poster), IRI meeting, June 2008, New York, NY (slides), EI fellows symposium, April 2008, New York, NY (slides)
(3) Models for public opinion of the death penalty.
I’m working with Andrew Gelman on fitting models for the public opinion of the death penalty, where our data comes from surveys given over the course of approximately 30 years, and we are most interested in modeling the change in public opinion by state over time.
Presentations:
August 2008, JSM, Denver, CO. “Modeling time series of death penalty public opinion” (slides pdf)
(4) Baseball fielding.
I’m working with Shane Jensen and Adi Wyner (Statistics dept., University of Pennsylvania) on fitting models to predict successful fielding plays in professional baseball, using high-resolution data on the exact location of batted balls, including flyballs, grounders, and liners. We have a great acronym for our system: SAFE (Spatial Aggregate Fielding Evaluation). Shane maintains a section of his website devoted to the project: SAFE (baseball fielding research).
_______________________________________________________________________________________________________________________
CONTACT
email: shirley@stat.columbia.edu
office: 927 School of Social Work Building (between 121st and 122nd on Amsterdam Ave., 1 floor below Statistics Department)