Bayesian Inference for Qualitative Research
Tasha Fairfield
The way we intuitively approach qualitative research is similar to how we read detective novels.
We consider different hypotheses to explain what happened—whether democratization in South Africa, or the death of Samuel Ratchett on the Orient Express—drawing on the literature we have read (e.g. theories of regime change or other Agatha Christie mysteries) and any other salient knowledge we have.
As we gather evidence and discover clues, we update our views about which hypothesis provides the best explanation—or we may introduce a new alternative that we think up along the way.
Bayesianism provides a logically rigorous and intuitive framework that governs how we should revise our views about which hypothesis is more plausible, given our relevant prior knowledge and the evidence that we find during our investigation.
Bayesianism is enjoying a revival across many fields, and it offers a powerful tool for improving inference and analytic transparency in qualitative research.
The principles we will cover in this module can be applied to single case studies (within-case analysis), comparative case studies (cross-case analysis) and multi-method research that draws on both qualitative evidence and quantitative data.
Causal Inference from Causal Models
Alan M. Jacobs
This module explores the use of causal models to design and implement qualitative and mixed-method empirical strategies of causal inference.
A great deal of recent methodological progress in the social sciences has focused on how features of a research design—such as randomization by the researcher or by nature—can allow for causal identification with minimal assumptions.
Yet, for many of the questions of greatest interest to social scientists and policymakers, randomization or its close equivalents are unavailable.
We are, in short, often forced to rely on beliefs about how the world works—that is, on models. Based on a book-in-progress by Macartan Humphreys and Alan Jacobs, this module will examine how we can engage in systematic model-based causal inference.
Specifically, we will explore how researchers can encode their prior knowledge in a probabilistic causal model (or Bayesian network) and an associated directed acyclic graph (DAG), use the model to make research design choices (including selecting cases and choosing observations), and draw inferences about causation at the level of both individual cases and populations, using both qualitative and quantitative data.
Choosing Spatial Units of Analysis
Hillel Soifer
TBA
Comparative Historical Analysis
Markus Kreuzer
We live in a constantly emerging world in which studying changes across time are just as crucial as analyzing differences across cases to understand our contemporary politics.
Comparative historical analysis (CHA) has long studied such historical changes and made important contributions to our understanding of how to use time to study the past.
It goes back to the 19th-century classics and shares more recently its ambitions with American political development, historical institutionalism and a long historical tradition in international relations.
These approaches all point out that time is to the past what grammar is to language and maps are to space: an essential tool of analysis.
This module explores three distinct contributions that CHA makes to our understanding of time.
First, it identifies distinct temporal building blocks that make time analytically tractable. Time scales specify how far into the past we look, chronologies specify the events we analyze and periodizations make historical contexts comparable.
These three building blocks constitute a historical notion of time that asks how different the past is from the present. CHA complements this historical notion of time with three elements of physical, clock-like time: duration, tempo and sequences which help to identify variations in the unfolding of the past.
Second, CHA uses these temporal building blocks for time spotting, that is foregrounding temporal and historical dynamics that many existing explanations background. In foregrounding time, CHA poses a series of macro-historical questions about the origins as well as continuous transformation of the state, political regimes, markets, war and global structures.
Third, CHA answers these macro-historical questions with the help of a range of causal mechanisms that explain the unfolding of historical processes through time. These mechanisms involve tipping points, diffusion, causal mechanisms, causal effects related to sequencing patterns, increasing and decreasing returns, and intercurrence (i.e. interaction among concurrent historical processes).
Overall, the model encourages students to spot elements of time that are hidden in their fields of research and explore how CHA can help them think about such elements more systematically, and thus enrich their analysis.
Ethnographic Methods
Sarah E. Parkinson and TBA
TBA
Fieldwork
Diana Kapiszewski, Lauren MacLean, Robert Mickey
This module introduces the fieldwork module sequence, considering the structure of the modules and presenting some of the overarching themes we will consider.
The module then begins to discuss the design, planning and execution of field research. We offer strategies for addressing the intellectual, social, emotional, health and logistical challenges that carrying out field research can involve.
Each session is conducted with the understanding that participants have carefully read the assigned materials. The instructors present key points drawing on the assigned readings, other published work on field research, and the experiences they and others have had with managing fieldwork's diverse challenges.
Interaction and discussion in small and large groups is encouraged.
Geographic Information Systems (GIS)
Jonnell Robinson
The module will introduce participants to GIS as a tool for qualitative research, present basic GIS terminology and concepts and the basic functions of ESRI's ArcGIS software suite, particularly those functions that are most commonly used by social scientists.
A subsequent session will explore basic visualization and analytical functions such as building and querying attribute tables, selecting map features and symbolizing data.
The module will also review the types and sources of data that are available for GIS users working in both data rich and data poor settings, the ethics of using mapping in research, how metadata can be used to communicate qualitative information, and data overlay analysis.
Other sessions will introduce open source geovisualization and analysis tools including Open Street Map, Google My Maps, and QGIS; demonstrate valuable data collection techniques for archival research, field work, participatory and community‐based mapping, as well as the availability and accessibility of spatial data through data repositories; and provide an overview of basic map design, integrating narrative and photos with GIS, and a discussion about why, how and where to further hone GIS skills.
Integrating Qualitative and Experimental Methods
Chris Carter, Charles Crabtree, Tesalia Rizzo Reyes, Guadalupe Tuñón
In this module sequence, we introduce natural and randomized experiments and discuss their strengths and limitations through a survey of recent examples from political science and economics. We introduce a common framework for understanding and assessing natural and randomized experiments based on the credibility of causal and statistical assumptions.
We discuss tools for developing and accessing experimental designs, such as instrumental variable analysis, sampling principles, power analysis, data collection do's and don'ts as well as a variety of robustness tests. We then discuss how to bolster the credibility of natural and randomized experiments in the design stage.
We will focus on the role of "ex-ante" approaches to improve the quality and transparency of research designs, such as the use of pre-analysis plans. The module incorporates applied research and practical advice, especially on how to conduct fieldwork, collect data, and analyze the logistics and ethics surrounding experiments.
We end the module by evaluating the promise and obstacles to the use of multi-method research in the analysis of natural and randomized experiments. We discuss how qualitative methods can help address some of the criticisms of experiments, as well as how experiments can bolster the inferences drawn from qualitative evidence.
Interpretation and History
TBA
Interpretive Methods
Lisa Wedeen, William Mazzarella
This two-module sequence provides students with an introduction to various modes of discourse analysis and ideology critique.
Students will learn to "read" texts while becoming familiar with contemporary thinking about interpretation, narrative, genre, and criticism.
In the first four sessions we will explore the following methods: Wittgenstein's understanding of language as activity and its practical relevance to ordinary language-use analysis; Foucault's "interpretive analytics" with hands-on exercises applying his genealogical method; and various versions (two sessions) of cultural Marxism—with specific attention to "ideology critique."
The last two classes will consider how anthropological discussions of participant observation can unsettle current versions of fieldwork in political science and, relatedly, how we might theorize practically forms of thought that appear to be paradoxical, nonsensical, or irrational.
Logic of Qualitative Methods
James Mahoney, Gary Goertz, Laura Garcia Montoya
This two-module (Modules 10 and 14) covers many classic and standard topics of qualitative methodology, with a special focus on how to write a qualitative dissertation or manuscript for publication as a book at an excellent university press.
We survey the key research design, case selection and theoretical issues that arise with such a project. The sessions use logic and set theory as a foundation for discussing and elucidating qualitative methods.
The individual topics for module 10 include a regularity theory of causality, large-N qualitative analysis (LNQA), and concepts. Module 14 focuses on process tracing.
After an introduction to process tracing, the module zooms in on two key topics: causal mechanisms and counterfactual analysis, central to process tracing.
Multi-Method Research
Jaye Seawright
This module looks at how to productively combine qualitative and quantitative methods when the overall goal is causal inference.
The module will be structured around a discussion of multi-method designs that use regression-type methods as the quantitative component of the causal inference.
For example, one session looks closely at combining case studies with regression, offering research designs for testing assumptions connected with measurement, confounding and the existence of a hypothesized causal path. Another will investigate case selection, asking how cases should best be selected from a larger population.
Participants will also investigate how multi-method research works in the context of random (or as-if random) assignment, exploring how to design case studies in conjunction with experimental or natural-experimental research.
Another session will consider what tools from statistics and machine learning can add to causal inferences based on process tracing.
Process Tracing and Typological Theories
Andrew Bennett
This module examines the inferential logic of process tracing, which is used extensively in qualitative case studies.
We identify Bayesian probability as one foundation for causal inference in process tracing, which entails assessing which hypothesis or theory provides the best explanation for the evidence at hand.
We will present practical advice for conducting process-tracing research as well as best practices for applying Bayesian reasoning in case study analysis.
In separate sessions, we will introduce causal graphs as a way to model causal processes and we will introduce the "Completeness Standard"—composed of causal graphs, event-history maps and invariant causal mechanisms—as a way to assess process tracing.
We will consider the extent to which successful execution of this standard supports valid claims of unit-level causal inferences.
Finally, we introduce typological theorizing as a way to address interaction effects and aid in selecting cases for process tracing, and we discuss examples of typological theories proposed in students' own work as well as in published research.
Re-thinking Small-N Comparisons
Nicholas Rush Smith
Why do we compare? Typically, in political science research, causal inference is taken as the primary goal.
Similarly, research that is generalizable to as many cases as possible tends to be valued more than research which can explain only a few.
This unit will push past these assumptions in two ways. First, it will provide logics for generalization not rooted in ideas of statistical generalizability or mechanical reproduction. Second, it will expand the goals of comparison from causal inference to alternative practices like creative redescription or conceptual development.
We will then explore how we can leverage strategies for rethinking comparison to address the practical challenges and unexpected discoveries that often upend pre-established research designs.
When a "crisis of research design" strikes, how can researchers cope with partially implemented data collection plans to still generate meaningful theoretical and empirical insights?
How can scholars salvage their research designs while maintaining methodological rigor?
Finally, we will critique short research designs that will be provided in advance.
Among other questions, we will ask ourselves: What kinds of claims can the author make with this research design and why? What are the limits on the kinds of claims they can make? How convincing is this research design? If you were on the selection committee of a funding agency, how would you rate this research design?