Dr Levi Wolf
Levi John Wolf’s website: ljwolf.org
Areas of Application
- Analysis of political behavior (elections, polarization)
- Spatial sorting, segregation, & spatial inequality
- Civic hacking & intervention
- Geographic data science
- Bayesian multilevel modeling (& hierarchical modeling in general)
- Models of spatial dependence
- Characterizing, modeling, & detecting spatial clusters, boundaries, & patterns
Example PhD Topics
- Spatial Model Criticism: Leamer’s “Let’s take the ‘con’ out of econometrics”suggests a research program full of model criticism. In spatial analysis, many of the methods used are applied in rote, with flow charts and checklists to ensure the “correct” application of techniques. In contrast, model criticism seeks to develop techniques to conduct percussive maintenance of our models, treating them as likely misspecified and underpowered. Adequate proposals here would focus on things like sensitivity to observations or hierarchies thereof, specification issues, tests to compare and select appropriate models or identify misspecification, or corrections to estimation techniques to adequately express uncertainty. Further, applications interested in focusing on improving the operational value of models, what the actual eventual estimate numbers represent would be helpful. Suggested readings would include E. Leamer, “Let’s take the con out of econometrics,” M. Wall “A close look at the spatial structure implied by the CAR and SAR models,” LeSage & Pace, Introduction to Spatial Econometrics.
- Boundary and Cluster Analysis: The analysis of disparity has often focused on relative comparisons between social groups or between predetermined communities. However, the geographies we experience do not tend to follow these administrative divisions. Methods to identify disparities, such as statistical boundary analysis (aka wombling) are used in epidemiology to detect disease clusters and barriers, but have not yet been applied in the analysis of society. Further, methods to identify communities, latent neighborhoods with social meaning to individuals, provides a richer notion of how individuals experience their geographies. Applications in this vein would be interested in the detection of latent spatial communities, neighborhoods, or clusters. They could be focused on characterizing which disparities between neighbors are so “extreme” as to be anomalous, and which might not be. It could focus on questions of residential sorting, political spatial & ideological polarization, income inequality, or other relevant subjects.
- Spatial Data Science: there is a wealth of machine learning techniques for large-scale data science that are applicable to social scientific inquiry. However, the relative advantage of social science (and a critical component of its social applicability) is its ability to provide interpretable insights into human behavior or social processes. As such, I would welcome any proposals on adapting or extending data science methods to social science problems (with a geographic component) where geographical science’s comparative advantage can be brought to bear. This may focus on examining whether geographic structures or thinking can improve typical data science analyses or can focus on the development of entirely novel techniques that bridge both geographic methods & data science techniques.
- Extending Spatial Multilevel Models: multilevel methods are a longstanding tool for spatial analysis. By default, however, they encode some fundamental restrictive assumptions about how things are related in space. Alternatives in the spatial econometric literature focus on processes and models where observations are dependent on their neighbors, regardless of group or hierarchical structures. A project in this vein would identify areas where methods to model spatial dependence can be integrated into hierarchical analysis. It would display a general interest in enriching spatial multilevel models with realistic structures to model spatial dependence. It could focus primarily on empirical application of these models, or be more strongly-focused in the statistical & analytical components. It might also encompass basic research into spatial econometric models that handle both heterogeneity & dependence.