New Methods - Speakers and Presentations
Sunday, October 15, 9:30-10:30 a.m.
"Comparing group effects in logit and probit models"
Presentation outline is available by clicking here.
J. Scott Long , Dr. J. Scott Long, Chancellor's Professor, Departments of Sociology and Statistics, Indiana University
Dr. J. Scott Long's research examines the interplay between women's work and family roles and its implications for physical health; human sexuality and sexual risk-taking in conjunction with The Kinsey Institute; stigma and mental health in a cross-national context; and methods for group comparisons in models with categorical outcomes. He is on the editorial board of Sociological Methodology, Sociological Methods and Research, Structural Equation Modeling and other journals. In 2002, he was awarded the 2002 Paul F. Lazarsfeld Memorial Award for Distinguished Contributions to Sociological Methodology. He is well-known for his contributions in the area of in regression models for categorical dependent variables.
Abstract: "Comparing group effects in logit and probit models"
Presentation outline is available by clicking here.
Abstract: Social scientists are often interested in assessing whether the effect of one variable on another is different in different groups of people. Gender, race, political party, country, economic region, experimental/control, and class are but a few examples where group comparisons of the effects of variables on some outcome are of fundamental importance. Extensions of the Chow (1960) test are often computed to compare groups, but this approach confounds the magnitude of the regression coefficients and the error variances in probit and logit models that assume an underlying latent variable, often yielding invalid conclusions. Indeed, Allison (1999) shows that these standard tests confound the magnitude of the regression coefficients and the variance of the error. Allison proposes a test that removes the effect of residual variation, but this test requires auxiliary information that is often unavailable. As an alternative, this talk proposes methods of group comparison based on predicted probabilities which can be applied to logit, probit and other types of regression models. We illustrate how this new method can be used and will describe software that we have developed to implement the approach.
Selected Publications:
Cheng, S. and J.S. Long, (forthcoming) Assessing measures of IIA for the multinomial logit model. Sociological Methods and Research.
Long, J. S. & Freese, J. (2005). Regression models for categorical dependent variables using Stata 2nd Edition. College Station TX: Stata Press.
Xu, J. and J.S. Long, 2005, Confidence intervals for predicted outcomes in regression models for categorical outcomes. The Stata Journal 5: 537-559.
Long, J. S. & Cheng, S. (2005). Regression Models for Categorical Outcomes. In Melissa Hardy and Alan Bryman (editors), Handbook of Data Analysis, Sage Publications.
Long, J. Scott. ( 1997). Regression models for categorical and limited dependent variables. Advanced Quantitative Techniques in the Social Sciences, Volume 7. Thousand Oaks CA: Sage Publications.
Bollen, K, & Long, J. S. (1993). Testing structural equation models. Newbury Park: Sage.
