Mar 12, 2014
The Toolkit for Weighting and Analysis of Nonequivalent Groups (TWANG) was first developed in 2004 by RAND researchers for the R statistical computing language and environment. The R version of the package contains functions for creating high-quality propensity score weights that can be used to estimate treatment effects with two or more treatment groups and time-varying treatments. The Shiny software development package allowed the TWANG project team to develop a menu-driven application that can be used to perform analyses using the TWANG package's suite of commands without requiring a user to learn R. This tutorial provides an introduction on how to use the TWANG Shiny app to estimate propensity score weights for two treatment groups when using observational data as well as to estimate treatment effect estimates using those propensity score weights.
The TWANG Shiny app can be used to compute the needed propensity score weights for an analysis and check the quality of the resulting propensity score weights by assessing whether they have good balancing properties. The TWANG Shiny app can also use the propensity score weights to estimate causal treatment effect estimates assuming key assumptions of methods hold (e.g., no unobserved confounding or overlap concerns exist in the data). It also showcases how to assess the sensitivity of those results to unobserved confounding. The tutorial demonstrates use of the TWANG Shiny app through an illustrative example and toy dataset. An updated version of the TWANG Shiny app was released in July 2022.