arXiv subjects

Amanda Kowalski

Publications and source records attributed to Amanda Kowalski.

5 recordsLinked to original sources

arXiv2020

Counting Defiers: Examples from Health Care

I propose a finite sample inference procedure that uses a likelihood function derived from the randomization process within an experiment to conduct inference on various quantities that capture heterogeneous intervention effects. One such quantity is the number of defiers---individuals whose treatment runs counter to the intervention. Results from the literature make informative inference on this quantity seem impossible, but they rely on different assumptions and data. I only require data on the cross-tabulations of a binary intervention and a binary treatment. Replacing the treatment variable with a more general outcome variable, I can perform inference on important quantities analogous to the number of defiers. I apply the procedure to test safety and efficacy in hypothetical drug trials for which the point estimate of the average intervention effect implies that at least 40 out of 100 individuals would be saved. In one trial, I infer with 95% confidence that at least 3 individuals would be killed, which could stop the drug from being approved.

stat.ME↗

arXiv2020

A Model of a Randomized Experiment with an Application to the PROWESS Clinical Trial

I develop a model of a randomized experiment with a binary intervention and a binary outcome. Potential outcomes in the intervention and control groups give rise to four types of participants. Fixing ideas such that the outcome is mortality, some participants would live regardless, others would be saved, others would be killed, and others would die regardless. These potential outcome types are not observable. However, I use the model to develop estimators of the number of participants of each type. The model relies on the randomization within the experiment and on deductive reasoning. I apply the model to an important clinical trial, the PROWESS trial, and I perform a Monte Carlo simulation calibrated to estimates from the trial. The reduced form from the trial shows a reduction in mortality, which provided a rationale for FDA approval. However, I find that the intervention killed two participants for every three it saved.

stat.ME↗

arXiv2020

Counting Defiers

The LATE monotonicity assumption of Imbens and Angrist (1994) precludes "defiers," individuals whose treatment always runs counter to the instrument, in the terminology of Balke and Pearl (1993) and Angrist et al. (1996). I allow for defiers in a model with a binary instrument and a binary treatment. The model is explicit about the randomization process that gives rise to the instrument. I use the model to develop estimators of the counts of defiers, always takers, compliers, and never takers. I propose separate versions of the estimators for contexts in which the parameter of the randomization process is unspecified, which I intend for use with natural experiments with virtual random assignment. I present an empirical application that revisits Angrist and Evans (1998), which examines the impact of virtual random assignment of the sex of the first two children on subsequent fertility. I find that subsequent fertility is much more responsive to the sex mix of the first two children when defiers are allowed.

stat.ME↗

arXiv2019

Censored Quantile Instrumental Variable Estimation with Stata

Many applications involve a censored dependent variable and an endogenous independent variable. Chernozhukov et al. (2015) introduced a censored quantile instrumental variable estimator (CQIV) for use in those applications, which has been applied by Kowalski (2016), among others. In this article, we introduce a Stata command, cqiv, that simplifes application of the CQIV estimator in Stata. We summarize the CQIV estimator and algorithm, we describe the use of the cqiv command, and we provide empirical examples.

econ.EM↗

arXiv2014

Quantile Regression with Censoring and Endogeneity

In this paper, we develop a new censored quantile instrumental variable (CQIV) estimator and describe its properties and computation. The CQIV estimator combines Powell (1986) censored quantile regression (CQR) to deal with censoring, with a control variable approach to incorporate endogenous regressors. The CQIV estimator is obtained in two stages that are non-additive in the unobservables. The first stage estimates a non-additive model with infinite dimensional parameters for the control variable, such as a quantile or distribution regression model. The second stage estimates a non-additive censored quantile regression model for the response variable of interest, including the estimated control variable to deal with endogeneity. For computation, we extend the algorithm for CQR developed by Chernozhukov and Hong (2002) to incorporate the estimation of the control variable. We give generic regularity conditions for asymptotic normality of the CQIV estimator and for the validity of resampling methods to approximate its asymptotic distribution. We verify these conditions for quantile and distribution regression estimation of the control variable. Our analysis covers two-stage (uncensored) quantile regression with non-additive first stage as an important special case. We illustrate the computation and applicability of the CQIV estimator with a Monte-Carlo numerical example and an empirical application on estimation of Engel curves for alcohol.

stat.ME↗