Dropout analysis: A method for data from Internet-based research and dropR, an R-based web app and package to analyze and visualize dropout.

Abstract

With Internet-based research, non-response such as lack of responses to particular items and dropout have become interesting dependent variables due to highly voluntary participation and large numbers of participants (Reips, 2000, 2002b). In this article, we develop and discuss the methodology of using and analyzing dropout in Internet-based research, and we present dropR, an R package and web service (web application) to analyze and visualize dropout. The web app was written in R using Shiny, a free software environment for statistical computing and graphics. Among other features, dropR turns input from datasets into accessible and publication-ready visual displays of dropout curves. It calculates parameters relevant to dropout analysis, such as chi-square values and odds ratios for points of difference, initial drop, and percent remaining in stable states. It provides Kaplan-Meier survival statistics and tests survival curve differences. With automated inferential components, it identifies critical points in dropout and critical differences between dropout curves for different experimental conditions (Kolmogorov-Smirnov and rho-family statistics) and produces related statistical copy. Requiring no programming knowledge, dropR is provided as a free web application at https://dropr.eu and for programmers as an R package (under a cost free general public license, GPL-3, https://cran.r-project.org/web/licenses/GPL-3 ) from researchers for researchers. All code and materials are openly available on GitHub ( https://github.com/iscience-kn/dropR ).