Evan Munro

I am a second-year PhD student in Economics at Stanford GSB, where I work with Guido Imbens and Susan Athey. I am interested in problems at the intersection of machine learning, econometrics, and mechanism design.

Previously, I received an MPhil in Economics from Oxford University, where my thesis was supervised by Anders Kock . I completed my B.A. at Columbia University, where I worked with Serena Ng on econometrics and big data.


Working Papers

Using Wasserstein Generative Adversial Networks for the Design of Monte Carlo Simulations (with Susan Athey, Guido Imbens, and Jonas Metzger)

Status: Revise and Resubmit

We discuss using Wasserstein Generative Adversarial Networks (WGANs) as a method for systematically generating artificial data that mimic closely any given real data set without the researcher having many degrees of freedom. We apply the methods to compare in three different settings twelve different estimators for average treatment effects under unconfoundedness.

Latent Dirichlet Analysis of Categorical Survey Responses (with Serena Ng)

Status: Revise and Resubmit

Existing methods that construct summary indices from survey data often treat the discrete variables as if they were continuous. We adapt hierarchical Bayesian methods used for text analysis to extract interpretable, low dimensional summaries from static and dynamic panels of survey data.

An Incentive-Compatible Draft Allocation Policy (with Martino Banchio)

Status: Submitted

To promote competition, sports leagues want to give the first round draft pick to teams with worse records each year. In response, many teams are accused of losing purposely in order to secure a better pick. We analyze draft pick allocation policy in the NBA using a theoretical model of team decision making. We prove that under the existing policy, it is not possible to have a lottery that is both redistributive and incentive-compatible. We design an alternative system that satisfies both redistribution and incentive compatibility.

Research Projects and Software

Deep Learning for Restaurant Choice

The state of the art methods in predicting consumer choice are Bayesian matrix factorization methods. This projects explores challenges and opportunities with using neural networks with embedding layers to predict restaurant choice from a large dataset of individual lunch choices in the Bay Area. Submitted as the final project for Stanford's Deep Learning class (CS230).

Identifying Side Products in Peptide Synthesis

Synthesis of long peptides via solid phase peptide synthesis can often lead to a number of side products. This simple R Shiny web application is meant to help determine which problematic side products might be appearing in the LC-MS of the crude cleaved peptide mixture. The utility implements a simple recursive solution of the subset sum problem. Originally written for members of the Peter Kim lab at Stanford.

Open Source Contributions

Package Language Contribution
grf R Boosted regression forests