Paul Allison's recent reserach on "Insiders, Outsiders and the Struggle for Consecration in Cultural Fields: A Core-Periphery Perspective," published in the American Socioloical Review is highlighted in the Penn Almanac,
Paul Allison and fellow collaborators, Gino Cattani, of New York University, and Simone Ferriani, of the University of Bologna, have published a study titled “Insiders, Outsiders, and the Struggle for Consecration in Cultural Fields: A Core-Periphery Perspective” according to a Penn News article. Their research focuses on how winning awards to gain recognition and success is dependent on who you know as much as who is judging.
Professor of Sociology
Ph.D., Sociology, University of Wisconsin-Madison, 1976
My principal contribution to this project is to provide training, advice and consultation on statistical analysis, and my methodological contributions related to all PARC Themes. My major areas of expertise are survival analysis, panel data analysis, categorical data analysis and missing data. Although randomized experiments are usually the preferred method for demonstrating causal relationships, they are usually not ethical or practical for answering the kinds of questions that most social scientists ask. So we have to settle for “second best” methods that have lots of potential pitfalls. I have been studying a collection of methods that enable us to get much closer to an experimental design. Known as “fixed effects methods,” these statistical techniques enable one to control for all stable characteristics of persons, regardless of whether we can measure those characteristics. They accomplish this by using each person as his or her own control. In some versions, these methods can also answer questions about the direction of causality: does X cause Y or does Y cause X? A 2006 paper in the New England J Medicine, with my former student and long-time collaborator, Nicholas Christakis (Harvard), is a significant example of the utility of this perspective. We highlight the strong and separable effects of the hospitalization and (perhaps) death of one spouse on the subsequent mortality risk of the other spouse. Part of the analysis drew on Cox survival models, where hazards of death can be estimated consequent to an event (hospitalization and/or death of index spouse), with controls for observed characteristics of individuals and couples. Analyses such as this assume that there are no unmeasured common characteristics of couples that might explain, for example, why two persons living together may sicken at the same time. There is a large literature (and complement of software) on fixed effects methods for cases in which there are repeated measures on individuals or couples over time as per the Medicare data that Christakis had assembled. Couples can be compared with themselves over time as a method of eliminating the effects of time-invariant unobserved couple characteristics; this is referred to in epidemiology as the case crossover method. A problem in the application of this method is that it assumes an independence of mortality risk from time. The details are too involved to go into here, but in a companion paper in Sociological Methodology 2006, we show how to turn the problem into one similar to that in case-control studies via logistic regression: estimates of the effects of one spouse’s death (or hospitalization) on that of the other are obtained by a conditional logistic regression stratifying by couple in which the outcome death (or non-death) at various segments of time is a predictor variable and the “treatment” (precipitating health event involving the other spouse) is the response. A series of simulations for this case-time-control method show excellent statistical properties (unbiasedness, coverage). The method well fits the application for which it was derived, but also generalizes to any number of sociological and demographic processes with non-repeatable end points and discrete treatments.
My other major research interest is statistical methods for handling missing data. Two new methods, multiple imputation and maximum likelihood, have been shown to be far superior to more traditional missing data methods. Nevertheless, they are still not widely used by social scientists. I hope to change that by making these methods easier to use and better understood. An example is my 2011 paper in the American J Sociology with Liana Sayers and Paula England, on differences between men and women in who initiates the departure from a marriage, and the relative effect of employment on this decision for each. It involved a great deal of innovative work with missing data and latent variables.
I have published 11 books and more than 75 peer-reviewed articles on these topics. I am a recipient of the ASA Paul Lazarsfeld Memorial Award for distinguished contributions in the field of sociological methodology and in 2010 I was elected as a Fellow of the American Statistical Association.