Making Statistical Comparisons: An Application Of The Bootstrap Technique In AIS Research

Abstract

In academic research, the classical approach to constructing confidence intervals and testing for significance has relied on the assumption of distribution normality. Nonparametric techniques do not require such assumptions, but have traditionally been difficult to implement. One such nonparametric technique is the bootstrap. In order to estimate the shape of a statistic’s sampling distribution, bootstrapping uses large numbers of repetitive computations rather than strong distributional assumptions and analytic formulas. Instead of imposing (by assumption) a shape on a statistic’s distribution, bootstrapping, through a resampling simulation, empirically estimates a statistic’s entire distribution. With the development and widespread use of the electronic spreadsheet, researchers are now able to easily and powerfully implement the bootstrap. This paper demonstrates how to create a bootstrapping template in a Quattro Pro for Windows spreadsheet package that will allow academic researchers to compare various instruments designed to measure user satisfaction.