Innovative Software for Prevention Scientists: SAS, R, Stata, and Web-Based Tools for Analysis and Design

This technology demonstration presents recent advances in SAS, R, STATA, and web-based software developed at The Methodology Center at Penn State. New and updated SAS procedures, R packages, STATA plug-ins, and free-standing web applets for latent class and latent transition analysis (LCA/LTA), causal inference, adaptive interventions, intensive longitudinal data, and fractional factorial designs will be showcased. All tools and corresponding users’ guides are available free-of-charge at methodology.psu.edu, and will be distributed on USB drives at the annual meeting.

Specifically, the following tools will be demonstrated: (1) an updated SAS macro (%LCA_Distal) for use with PROC LCA that estimates the association between latent class membership and a distal outcome that now includes standard errors; (2) a new SAS macro (%GBM_binary) that produces propensity scores for causal inference using generalized boosted modeling by calling an R package (twang.R); (3) an updated STATA plug-in to conduct LCA; (4) a new R package (qlaci.R) to analyze data from sequential, multiple assignment, randomized trials (SMARTs) in order to inform development of adaptive interventions; (5) an expanded suite of SAS macros to analyze intensive longitudinal data using the time-varying effects model (TVEM) that allows the estimation of both time-invariant and time-varying effects of predictors on an intensively measured outcome; and (6) a new utility to randomly assign participants to conditions in fractional factorial designs.   

In addition, information will be available about other software created and maintained by The Methodology Center at Penn State. For example, this includes SAS example code for conducting multiple imputation with PROC LCA, SAS and R example code for conducting propensity score analysis, and SAS macros (%RelativeCosts1 and %FactorialPowerPlan) to calculate the costs of multiple-factor experimental designs (for use with the Multiphase Optimization Strategy [MOST]). Demonstrations and information will be presented in the context of empirical data on the etiology or prevention of substance use and related behaviors. Attention will be given to practical implementation of the statistical methods using the demonstrated software, as well as recent features that have been incorporated into the software.