Ph.D. Dissertation Research

Details forthcoming!

 

Graduate Research

Interference in Field Experiments:

For my external research rotation under the Big Data Social Science fellowship, I worked with Dr. Bruce Desmarais from the Political Science department at Penn State. When SUTVA is violated via networked social interaction, treatment effects spread to control units through the network structure. We reviewed recently developed methods that can be used to account for interference in the analysis of data from field experiments on state legislatures. These methods required the researcher to specify a spillover model according to which legislators influence each other, and specify the network through which spillover occurs. We discuss these and other specification steps in detail. An earlier version of this paper was presented at the 9th Annual Political Networks Conference 2016  and the Joint Statistical Meeting 2016.

(Phadke, S., & Desmarais, B., “Considering Network Effects in the Design and Analysis of Field Experiments on State Legislatures.” State Politics & Policy Quarterly, 19(4):451-473, 2019.)

Data Science for the Public Good Fellowship:

During the summer of 2017, I was one of the four graduate fellows in this program organized by the Social and Data Analytics lab of Virginia Tech. In these ten weeks, I worked on four separate projects under the guidance of faculty at the lab, collaborating with other graduate fellows, and guiding undergraduate students.

• Open Source Software (OSS): The National Science Foundation (NSF) is interested in understanding the scope and impact of Open Source Software. Our team scraped data from two websites in an attempt to understand the universe of OSS, and summarized the development, presence, and purpose of OSS based on the data collected. [Final poster]
• Arlington County Department of Human Services (DHS): DHS stores client data for the various services it provides in different databases. Linking these data across services faces issues such as availability of all information and data entry errors. This makes it difficult to answer fundamental questions such as, how many unique clients are served by DHS in a single day?, and what are the demographics of those clients? Our team implemented a probabilistic linkage algorithm to improve matching records for every individual in the system. Here is a Git repository maintained by my co-fellow, Zarni Htet. [Final poster]
• Arlington County Police Department (ACPD): ACPD is interested in finding out how various events; political, social, weather-related, and others impact crime in Arlington, in order to take measure apriori and prepare themselves. We explored these phenomena geographically and temporally. In addition to data visualizations, we used a spatiotemporal log-Cox point process to model crime rates while accounting for dependence across time. [Final poster]
• Alexandria County Equity Group: The City of Alexandria is striving towards combating inequity in their community. We conducted a literature review to better understand how equity is defined and measured. Our goal was to explain how these definitions and indicators of inequity manifest themselves in Alexandria. We provided recommendations to the city about forming an actionable definition of equity, and collecting the most pertinent data to measure it. [Final poster]

 

Pre-graduate-school Research

Causal Inference:

During my research assistantship at the Applied Statistics and Computing Lab (Indian School of Business), I had a chance to work with two senior faculty members at the school. We explored counterfactual methods for causal inference and how they can be applied to social science research. We were the first set of researchers in the school to study these methods. We conducted an extensive literature review to familiarize ourselves with the Rubin Potential Outcomes Framework, Judea Pearl’s DAGs methods, and other methods for causal inference. I made 12 presentations to the faculty, and was a teaching assistant for the Causal Inference course taught to the students of FPM (Fellow Program in Management) and other researchers at the school.

Studying Indian Family Businesses:

The goal of this study was to study the performance of Family businesses in India relative to other businesses such as Multi-national Corporations, Public Sector Units, and Non-family businesses, after economic liberalization in India in the year 1992. The longitudinal dataset comprised of listed companies. We worked on a panel data model suitable for explaining the Profit After Tax (PAT) based on important financial variables and other relevant characteristics. We worked with the primary researchers, knowledgeable in the subject matter, and authored a working paper with Professor Kavil Ramachandran and Professor Bhimasankaram Pochiraju, from Indian School of Business, Hyderabad, India.