Welcome!

Hello, I'm Jason Morgan, a Data Scientist at Nationwide and visiting scholar in the Department of Political Science at The Ohio State University.

My Work

I use a number of modeling and machine learning techniques to solve difficult problems. I have recently been engaged in a number of projects, including:

• Designing and analyzing experiments focused on improving sales conversions and business operations;
• Developing automated computer vision models for image segmentation, feature detection, and optical character recognition (using OpenCV and Tesseract); and
• Building predictive recommendation models for online sales engines (employing hierarchical and PAM clustering, gradient boosting, and Bayesian hierarchical modeling).

That's my day job. I also maintain an academic research agenda, focused on models for network analysis, particularly latent space models and advanced exponential random graph models, as well as the computational and implementational aspects of estimating these models. I have also done research on causal inference in the context of natural experiments (instrumental variables and regression discontinuity designs).

Education

I earned my Ph.D. at OSU, focusing on applied political methodology and comparative politics. Prior to being at OSU, I earned an M.A. in political science from Boston College and my B.A. in economics and German studies from Lewis & Clark College.

Research

Published Research

• How Strong is Strong Enough? Strengthening Instruments through Matching and Weak Instrument Tests (with Luke Keele). Annals of Applied Statistics. Abstract.

  In a natural experiment, treatment assignments are made through a haphazard process that is thought to be as-if random. In one form of natural experiment, encouragement to accept treatment rather than treatments themselves are assigned in this haphazard process. This encouragement to accept treatment is often referred to as an instrument. Instruments can be characterized by different levels of strength depending on the amount of encouragement. Weak instruments that provide little encouragement may produce biased inferences, particularly when assignment of the instrument is not strictly randomized. A specialized matching algorithm can be used to strengthen instruments by selecting a subset of matched pairs where encouragement is strongest. We demonstrate how weak instrument tests can guide the matching process to ensure that the instrument has been sufficiently strengthened. Specifically, we combine a matching algorithm for strengthening instruments and weak instrument tests in the context of a study of whether turnout influences party vote share in US elections. It is thought that when turnout is higher, Democratic candidates will receive a higher vote share. Using excess rainfall as an instrument, we hope to observe an instance where unusually wet weather produces lower turnout in an as-if random fashion. Consistent with statistical theory, we find that strengthening the instrument reduces sensitivity to bias from an unobserved confounder.

• Modeling Unobserved Heterogeneity in Social Networks with the Frailty Exponential Random Graph Model (with Jan Box-Steffensmeier and Dino Christenson). Forthcoming in Political Analysis. Abstract.

  In the study of social processes, the presence of unobserved heterogeneity is a regular concern. It should be particularly worrisome for the statistical analysis of networks, given the complex dependencies that shape network formation combined with the restrictive assumptions of related models. In this paper, we demonstrate the importance of explicitly accounting for unobserved heterogeneity in exponential random graph models (ERGM) with a Monte Carlo analysis and two applications that have played an important role in the networks literature. Overall, these analyses show that failing to account for unobserved heterogeneity can have a significant impact on inferences about network formation. The proposed frailty extension to the ERGM (FERGM) generally outperforms the ERGM in these cases, and does so by relatively large margins. Moreover, our novel multilevel estimation strategy has the advantage of avoiding the problem of degeneration that plagues the standard MCMC-MLE approach.

• Web Timing Attacks Made Practical (with Tim Morgan). Blackhat 2015. Abstract. Video.

  This paper addresses the problem of exploiting timing side channels in web applications. To date, differences in execution time have been difficult to detect and to exploit. Very small differences in execution time induced by different security logics, coupled with the fact that these small differences are often lost to significant network noise, make their detection difficult. Additionally, testing for and taking advantage of timing vulnerabilities is often hampered by the tools available. To that end, we perform a thorough Monte Carlo comparison of several statistical techniques meant to identify the existence of differences in computation time in remote web applications. We then implement a tool that allows penetration testers to more thoroughly identify potential exploits

Manuscript

• Inferential Network Analysis (with Skyler Cranmer and Bruce Desmarais). Under contract with Cambridge University Press.

Software

• dynnet. An R package providing an alternative implementation latent space models for static and dynamic network models. Meant as a test-bed for exploring new ideas.
• boolean3. A reimplementation of the boolean package in R. Provide multiprocessor support, improve performance, extended capabilities.
• iv_sens. An implementation of Rosenbaum's instrumental variable sensitivity analysis for causal inference. Available in the rbounds package for R (with Luke Keele).