37: Reintroduction of the bdots Package for Bootstrapped Differences in Timeseries

Music City Center 

In 2018, Seedorff et al., introduced the bdots package to CRAN, an implementation of the bootstrapped differences in time series (BDOTS) methodology first introduced by Oleson et. al 2017. Originally imagined in the context of the Visual World Paradigm, BDOTS presents a novel method for controlling Type I error rates when identifying differences between densely sampled time series. This poster presents two critical changes to the bdots package: first, it corrects an issue in the original method that drastically underestimates the Type I error rate; an examination of this, along with the power of proposed alternatives is presented. Second, the entire package has been rewritten to accommodate a number of important functions, including paired testing and the ability to define arbitrary curve fitting functions to the observed data. Examples of syntax and the utility of the added functionality are also presented.

bootstrap

time series

R

permutation

functional programming

VWP 

Section for Statistical Programmers and Analysts