OBME 2185 - Computational Methods for Social Science with Python

This course introduces students to the logic and methods of social classification. Social scientists often characterize social processes as nuanced, many-dimensional, and even irreducibly complex. Typologies, taxonomies, clusters, and other kinds of classification have a great deal of appeal for answering questions about complex objects of study: we can gain insight into seemingly intractably complicated social processes by first imposing simpler analytic structure and then allowing this structure to guide our inquiry.

Developments in computer science and machine learning over the past three decades have transformed the practice of social classification. It has never been easier to apply complex classification schemes to large amounts of data. At the same time, the ease of applying complex classification methods combined with qualitative changes in the kinds of data available for research both introduce new complexities into the research process. If not handled carefully, these complexities can limit the usefulness and generality of our results. This course will teach students how to recognize situations in which classification may (not) be appropriate or useful; how to apply classification algorithms and related machine learning tools appropriately; and how to address common failure modes for applications of classification in the social sciences.

Topics to be covered include:

• how to translate research questions into classification problems; how to determine whether classification is an appropriate method for a given question; and how to evaluate the robustness and validity of a classification
• a framework for methods classifying a set of units (e.g., people, countries, life sequences, words, documents, images) based on patterns in data describing their common characteristics; how to use R to implement these methods
• relationships between classification and related pattern analysis techniques (e.g., clustering, regression, partitioning, mixed-membership modeling)

Learning outcomes

Students will:

• Understand the logic and rhetoric of classification in the social sciences.
• Be able to apply classification algorithms to problems of interest in the social sciences using standard packages in R.
• Become informed and critical consumers of results in quantitative social research that rely on classification.

• One prior course in statistical data analysis is required, preferably with an emphasis on linear regression.
• Prior experience working in R is not required, but strongly recommended. Students without prior R experience will be asked to invest additional time in R practice outside of class during the first few weeks of the semester.

• Reading assignments. Students are expected to prepare for seminar each week by completing the assigned reading(s).
• Seminar attendance and participation. Students are expected to participate actively in seminar discussions about the benefits and drawbacks of classification methods.
• Project proposal, pilot analysis, and final presentation. Students will author a proposal for a project that uses classification methods to answer a research question of their choice; conduct a pilot analysis that applies a classification method to relevant data and evaluates the results; and present their findings to their peers in class at the end of the semester. You will also be asked to schedule a one-on-one meeting with me mid-semester to discuss your plans.

• In Class Presence: 2 hours a week / 24 hours a semester
• Reading and Preparation for Class: 1 hour x 12 weeks = 12 hours a semester
• Research and Writing for Individual Assessments: 3 hours x 1 weeks = 36 hours a semester