Using full matching to estimate causal effects in nonexperimental studies: Examining the relationship between adolescent marijuana use and adult outcomes.

Abstract

Matching methods such as nearest neighbor propensity score matching are increasingly popular techniques for controlling confounding in nonexperimental studies. However, simple k:1 matching methods, which select k well-matched comparison individuals for each treated individual, are sometimes criticized for being overly restrictive and discarding data (the unmatched comparison individuals). The authors illustrate the use of a more flexible method called full matching. Full matching makes use of all individuals in the data by forming a series of matched sets in which each set has either 1 treated individual and multiple comparison individuals or 1 comparison individual and multiple treated individuals. Full matching has been shown to be particularly effective at reducing bias due to observed confounding variables. The authors illustrate this approach using data from the Woodlawn Study, examining the relationship between adolescent marijuana use and adult outcomes. Keywordslongitudinal studies; long-term consequences; observational study; propensity score; substance use In nonexperimental studies, researchers are often interested in examining the effect of some event or treatment (e.g., substance use) on an outcome (e.g., educational attainment). This is done by comparing individuals who experienced that event or treatment (e.g., substance users) to individuals who did not (e.g., nonusers). In experimental studies with random assignment, treatment and control groups are similar on all background characteristicsobserved and unobserved-as a consequence of the randomization, allowing for straightforward comparison of outcomes. In contrast, in nonexperimental studies, the treatment and comparison individuals may differ significantly on background characteristics -some that are observed and others that may be unknown. For example, substance users and nonusers are likely to be different on characteristics such as family history of drug use as well as on individual behaviors such as aggression. Thus, any difference in outcomes between the two groups may be due to these background covariates or to the treatment itself Author ManuscriptAuthor Manuscript Author ManuscriptAuthor Manuscript (i.e., substance use). The question then is how best to compare substance users with nonusers to clearly separate the effects of substance use from any of these other differences in background characteristics.Matching methods, such as nearest neighbor propensity score matching, are increasingly popular techniques for controlling for observed confounding variables when estimating causal effects in nonexperimental studies. The goal of matching methods is to ensure that the distributions of observed covariates in the treatment and comparison groups are similar, replicating what would have occurred had the treatment been randomly assigned, at least with respect to the observed covariates. Although regression has often been used to adjust for background differences and estimate causal effects in nonexperimental studies, it relies heavily...

Early Life Predictors of Adult Depression in a Community Cohort of Urban African Americans

“…1,56,57 Instead, Woodlawn men have rates of depression (16.2 %) only slightly lower than Woodlawn women (18.8 %), demonstrating the magnitude of the problem of depression for African American urban men. Woodlawn men also have high rates of crime, homelessness, substance use, and unemployment, [58][59][60][61] all of which have been found to be associated with depression.…”

Estrategia para la inferencia causal y planificación de estudios observacionales en las ciencias sociales: el caso de Chaitén post erupción del 2008

“…Si la razón de tratados y controles dentro de cada bloque está muy desbalanceada, dicho problema puede introducir sesgo en las estimaciones (Stuart y Green 2008). Tal como sugiere Hansen (2004), Full Matching permite solucionar este problema a través de la imposición de restricciones al número de tratados y controles que habrá dentro de cada bloque.…”

“…Los análisis de balance fueron realizados con el paquete de R RItools (Bowers et al 2016 Las diferencias estandarizadas complementan al test ómnibus en la medida que ellas permiten evaluar el balance para cada covariable por separado. Diferencias menores de 0,25 en términos absolutos son habitualmente consideradas como un balance aceptable (Stuart y Green 2008). Los resultados de la Tabla 1 en la condición pre-matching indican que el balance sería aceptable para la mayoría de las covariables salvo el género, el indicador de si el encuestado llegó en 2011 o después a Chaitén y, sobre todo, las variables de estatus socioeconómico (ingreso y nivel educacional).…”

Long-term Effects of Adolescent Smoking on Depression and Socioeconomic Status in Adulthood in an Urban African American Cohort

“…48 Applications of propensity score methods in examining longitudinal effects in observational research include, but are not limited to, studies that looked at developmental trajectory and substance use. 39,[49][50][51][52] The propensity score matching method was used to account for early-life or childhood covariates in relation to adolescent smoking and to obtain a matched sample. For propensity score matching, all analyses were conducted with STATA statistical software.…”