Kari Lock Morgan

Balancing Covariates via Propensity Score Weighting

ABSTRACT Covariate balance is crucial for unconfounded descriptive or causal comparisons. However, lack of balance is common in observational studies. This article considers weighting strategies for balancing covariates. We define a general class of weights—the balancing weights—that balance the weighted distributions of the covariates between treatment groups. These weights incorporate the propensity score to weight each group to an analyst-selected target population. This class unifies existing weighting methods, including commonly used weights such as inverse-probability weights as special cases. General large-sample results on nonparametric estimation based on these weights are derived. We further propose a new weighting scheme, the overlap weights, in which each unit’s weight is proportional to the probability of that unit being assigned to the opposite group. The overlap weights are bounded, and minimize the asymptotic variance of the weighted average treatment effect among the class of balancing weights. The overlap weights also possess a desirable small-sample exact balance property, based on which we propose a new method that achieves exact balance for means of any selected set of covariates. Two applications illustrate these methods and compare them with other approaches.

Rerandomization to Balance Tiers of Covariates

When conducting a randomized experiment, if an allocation yields treatment groups that differ meaningfully with respect to relevant covariates, groups should be rerandomized. The process involves specifying an explicit criterion for whether an allocation is acceptable, based on a measure of covariate balance, and rerandomizing units until an acceptable allocation is obtained. Here, we illustrate how rerandomization could have improved the design of an already conducted randomized experiment on vocabulary and mathematics training programs, then provide a rerandomization procedure for covariates that vary in importance, and finally offer other extensions for rerandomization, including methods addressing computational efficiency. When covariates vary in a priori importance, better balance should be required for more important covariates. Rerandomization based on Mahalanobis distance preserves the joint distribution of covariates, but balances all covariates equally. Here, we propose rerandomizing based on Mahalanobis distance within tiers of covariate importance. Because balancing covariates in one tier will in general also partially balance covariates in other tiers, for each subsequent tier we explicitly balance only the components orthogonal to covariates in more important tiers.

Measuring Geographic "Hot Spots" of Racial/Ethnic Disparities: An Application to Mental Health Care.

Abstract:This article identifies geographic “hot spots” of racial/ethnic disparities in mental health care access. Using data from the 2001–2003 Collaborative Psychiatric Epidemiology Surveys(CPES), we identified metropolitan statistical areas(MSAs) with the largest mental health care access disparities (“hot spots”) as well as areas without disparities (“cold spots”). Racial/ethnic disparities were identified after adjustment for clinical need. Richmond, Virginia and Columbus, Georgia were found to be hot spots for Black-White disparities, regardless of method used. Fresno, California and Dallas, Texas were ranked as having the highest Latino-White disparities and Riverside, California and Houston, Texas consistently ranked high in Asian-White mental health care disparities across different methods. We recommend that institutions and government agencies in these “hot spot” areas work together to address key mechanisms underlying these disparities. We discuss the potential and limitations of these methods as tools for understanding health care disparities in other contexts.

Taking a Chance in the Classroom: Making the Old New Again

making the old new Again The harmful effects of high blood-lead levels in children have been known and reported for more than 35 years, yet related headlines are still common. Using the internet it is not hard to find recent reports to convince students that lead poisoning in children, a problem discovered long ago, is an important modern-day global health concern. Excerpts from three postings within the last year are included here:

Taking a Chance in the Classroom: Exploring Google's Transparency Report

In October of 2011, Google released its transparency report, revealing the number of requests received from government agencies and courts around the world for content removal from web search results or other Google-owned sites such as YouTube and Blogger, as well as data on users involved in criminal cases. Google summarizes why they chose to release this information in their report: “Transparency is a core value at Google. As a company, we feel it is our responsibility to ensure that we maximize transparency around the flow of information related to our tools and services. We believe that more information means more choice, more freedom, and ultimately more power for the individual.” Using the original data released by Google and additional countrylevel variables such as population, Internet access, and indicators of development and democracy, we show how teachers and students can investigate relationships between countries’ data request behavior, Google’s compliance rate, and other country characteristics. This data set is timely and likely to be of interest to students at all levels, especially at a time when Internet censorship and data privacy are such hot topics. It is also a rich data set for thorough exploratory data analysis and illustrating problems with missing and censored data. Standard inferential techniques are not really appropriate for analyzing these data since the 42 countries included in the data set do not constitute a representative Exploring Google’s transparency report