Machine Learning Predictions as Regression Covariates
- 11 November 2020
- journal article
- research article
- Published by Cambridge University Press (CUP) in Political Analysis
- Vol. 29 (4) , 467-484
Abstract
In text, images, merged surveys, voter files, and elsewhere, data sets are often missing important covariates, either because they are latent features of observations (such as sentiment in text) or because they are not collected (such as race in voter files). One promising approach for coping with this missing data is to find the true values of the missing covariates for a subset of the observations and then train a machine learning algorithm to predict the values of those covariates for the rest. However, plugging in these predictions without regard for prediction error renders regression analyses biased, inconsistent, and overconfident. We characterize the severity of the problem posed by prediction error, describe a procedure to avoid these inconsistencies under comparatively general assumptions, and demonstrate the performance of our estimators through simulations and a study of hostile political dialogue on the Internet. We provide software implementing our approach.Keywords
This publication has 14 references indexed in Scilit:
- Improving Ecological Inference by Predicting Individual Ethnicity from Voter Registration RecordsPolitical Analysis, 2016
- Anti-Americanism and Anti-Interventionism in Arabic Twitter DiscoursesPerspectives on Politics, 2015
- Political Ideology Detection Using Recursive Neural Networks2014
- How Censorship in China Allows Government Criticism but Silences Collective ExpressionAmerican Political Science Review, 2013
- How Words and Money Cultivate a Personal Vote: The Effect of Legislator Credit Claiming on Constituent Credit AllocationAmerican Political Science Review, 2012
- A Method of Automated Nonparametric Content Analysis for Social ScienceAmerican Journal of Political Science, 2009
- Use of force and civil–military relations in Russia: an automated content analysisSmall Wars & Insurgencies, 2009
- Missing-Data Methods for Generalized Linear ModelsJournal of the American Statistical Association, 2005
- The New Videomalaise: Effects of Televised Incivility on Political TrustAmerican Political Science Review, 2005
- Regression with a binary independent variable subject to errors of observationJournal of Econometrics, 1973