- Robert J. Franzese, University of Michigan
Cross-unit (i.e., "spatial") interdependence is ubiquitous throughout the social sciences. Events or outcomes in one observational unit are often related to similar occurrences in other observational units. This is the case for such diverse phenomena as disturbances and conflicts within and among nations; consumer and producer choices in markets; opinions and behavior in societies; crime, health, and environmental outcomes; voting by citizens in elections or by legislators in legislatures; and policies in political jurisdictions. In such contexts, "standard" statistical methods (which assume independent observations) are inappropriate. This workshop introduces strategies appropriate for interdependent observations, using spatial and spatiotemporal models of interdependent continuous and limited outcomes.
The main objective of the workshop is to show how such spatial, cross unit, interdependence can be incorporated into empirical analysis. Course participants will learn how to: diagnose spatial-correlation patterns; estimate spatial-regression models; distinguish between different sources of spatial correlation (common exposure, contagion, and selection); and calculate and present the spatial and spatiotemporal effects that empirical models which incorporate interdependence imply. Methods to be covered include: measures of spatial association; instrumental-variable and maximum-likelihood estimators for regression models with spatial interdependence; multiple-spatial-lag models; spatial interdependence in models with limited and qualitative dependent-variables; and models for coevolutionary processes.
Prerequisites: Participants in this course should have some familiarity with linear regression and models for qualitative/limited dependent variables (e.g., logit, probit, etc.). This workshop does not assume any prior knowledge of, or experience with, spatial statistics.
NOTE: This course does not cover geographical information systems (GIS) or geocoding of spatial data, except insofar as the methods covered are frequently applied to such data. Also, for purposes of this course, the term "spatial" means "among spatial units." The correlation and/or dependence in question may or may not be geographically based.
Fee: Members = $1700; Non-members = $3200
Location: ICPSR -- Ann Arbor, MI
Date(s): July 17 - July 21
Time: 9:00 AM - 5:00 PM