Adam Lauretig

Data Scientist

Biography

I am a Data Scientist in New York City.

In the summer of 2019, I received my Ph.D. in Political Science at The Ohio State University, with a focus on political methodology and foreign policy decision-making. In my dissertation, I derived a new model, Bayesian Word Embeddings with Automatic Relevance Determination priors, which allows users to measure the meaning of individual words, and incorporates an estimate of uncertainty in the embeddings. I also developed an inference strategy to identify embeddings, so embeddings can be used in regression models for hypothesis testing. I used this model to investigate the determinants of threat perception during the Cold War among elites, drawing on the recently digitized Foreign Relations of the United States.

I’ve written a few tutorials, one for R users who want to use the Unity cluster at Ohio State, and another for R users who want an introduction to word2vec in Gensim.

Lately, I’ve been giving the following talks:

A Common Model, Separated by Two Disciplines: Bayesian Factorization Machines with Stan and R

Identification and Inference in Bayesian Word Embeddings

Interests

Political Science
Political Methodology
Text as Data
Bayesian Statistics
Causal Inference

Education

Ph.D. in Political Science, 2019

The Ohio State University
B.A. in Political Science, 2013

Grinnell College

Recent Publications

AM Lauretig. Identification and Inference in Bayesian Word Embeddings. In the Proceedings of the Third Workshop on NLP and Computational Social Science. 2019., 2019.

Preprint PDF Code Dataset

A Sacks, J Pierskalla, AM Lauretig, A Rosenberg. Mapping Indonesia's Civil Service. Policy Report. World Bank Group., 2018.

PDF

AM Lauretig, BF Braumoeller. Statistics and International Security. In The Oxford Handbook of International Security, Oxford University Press., 2018.

Preprint PDF Dataset

Recent Posts

Feb 10, 2019 Text as Data

Tokenizing on Stopwords

Introduction Recently, I came across the idea that you can get relevant keywords for word2vec by tokenizing a corpus on stopwords, in addition to standard punctuation (found via). This seemed like a really cool unsupervised way of capturing (hopefully!) relevant phrases. I was intrigued. A brief note: “tokenizing” refers to splitting a document into words or phrases based on a pre-defined set of rules. The most common way to do this is by splitting on spaces and “end-of-sentence” punctuation (ex: “!

Jan 31, 2019 Data Cleaning

Monthly Militarized Interstate Dispute (MID) Panel

Just a brief note. For my dissertation, I needed to create a monthly panel of militarized interstate dispute (MID) data, which proved to be more difficult than initially anticipated. To “be kind to my future self,” I wrote up what I did on a gist, and I’ll also copy the code below. If you find yourself in a similar situation, I hope this helps! # Code to create a monthly, directed dyadic panel of # Militarized Interstate Dispute data, with directional initiation # uses cshapes to create monthly list of all dyads, then uses data.

Nov 3, 2018 Causal Inference

Generalized Propensity Score Weighting

Generalized Propensity Score Weighting Adam Lauretig 2018-11-03 Introduction I’m writing this gist to better understand both the generalized propensity score, and marginal structural models, and especially, their intersection. In this, the goal is to estimate how a real-valued treatment can be estimated and understood for a time varying effect. Here, I denote treatment A at time t as At, and covariates X for individual i at time t as Xi, t.

Teaching

I was the instructor of record for POL4782: Data Analysis in Political Science II in Spring 2018, the syllabus is available here. I taught students regression modeling using ordinary least squares (OLS), and generalized linear models (GLMs) using Maximum Likelihood Estimation. The Rmarkdown template I had my students use for problem sets and materials for getting acquainted with Rmarkdown are available here. I hope you find them helpful!

I was the graduate teaching assistant for the first course in the graduate quantitative methods sequence in Fall 2018. I was responsible holding recitations for teaching students the R software package, the materials I used are available here.