Propensity score based data analysis using nonrandom

For some time, propensity score based methods have been frequently applied in the analysis of data from observational studies. The propensity score is the conditional probability of a certain treatment or exposure given patient’s covariates. Propensity score methods are used to eliminate baseline imbalances in covariate distributions between treatment or exposure groups and permit to estimate marginal effects. The package nonrandom is a tool for a comprehensive data analysis using stratification and matching by the propensity score. Several functions are implemented, starting from the selection of the propensity score model up to estimating propensity score based treatment or exposure effects. Before estimating the propensity score, relative.effect() permits to investigate the extent to which a covariate is confounding the treatment or exposure effect. This measure may support the decision to include a covariate in the propensity score model. pscore() estimates the propensity score and provides all information about the model. Stratification and matching by the propensity score are implemented in ps.makestrata() and ps.match(), respectively. To check the balance of covariate distributions between treatment or exposure groups, ps.balance() tests the distributions using statistical tests or standardized differences and dist.plot() allows for a graphical balance check. Finally, propensity score based estimators for the treatment or exposure effect can be determined by ps.estimate(). It also provides a comparison to regression based estimates alternatively used. All functions can be applied separately as well as combined. Additionally, it is possible to apply all functions repeatedly to decide which analysis strategy is the most suitable one. There are two data examples to illustrate the application of nonrandom. In the first data example, quality of life is investigated in breast cancer patients in an observational treatment study of the German Breast Cancer Study Group (GBSG). The second data example deals with lower respiratory tract infections (LRTI) in infants and children in the observational study Pri.DE (Pediatric Respiratory Infection, Deutschland) in Germany.