Advanced Research Methods
This is one of four required courses in the core curriculum of the LBJ School’s Ph.D. Program in Public Policy. Its purpose is to explore unifying themes in applied social and policy research and to lay foundations for designing and carrying out solid policy inquiry. Central to this will be developing the following key perspectives:
Aligning substantive theory with appropriate methods.
Building a framework for causal research that spans both qualitative and quantitative methods.
Learning to use directed acyclic graphs to model causal relations.
Reflecting on the limits/threats to causal inference and to internal and external validity.
Understanding the relative virtues of design-based and model-based inference.
Appreciating the power of Bayesian methods to “borrow strength” in the way it integrates theory and empirical evidence.
Using modern computer-intensive methods to reflect critically on conventional textbook statistics.
Marshaling and managing information for conducting policy research (survey methods, organizing/merging secondary data, etc).
To make these learning goals concrete, the course organizes the above themes around the principles of the microeconometric evaluation of social programs and policies. Students will be encouraged to apply these principles to their own specific dissertation research arena. Students willl be evaluated in terms of the quality of class participation, in-class presentations, fortnightly take-home exercises, and performance on two special take-home assignments.
The course is tailored to the LBJ School's Ph.D. program, but doctoral students from other departments and a limited number of qualified second-year LBJ master’s students are welcome to enroll with the instructor's consent (please e-mail instructor for consent). The prerequisite is the equivalent of two semesters of graduate-level statistics with a comfortable command of linear regression analysis and a decent exposure to limited dependent variable regression (logit, probit, or poisson maximum likelihood models).