Bootstrapped Confidence Intervals as an Approach to Statistical Inference

Abstract

Confidence intervals are in many ways Keywords: bootstrap; confidence interval; hypothesis test; resampling; statisticsResearch in the organizational sciences makes frequent use of statistical inferences. These inferences often take the form of null hypothesis tests based on mathematical probability theory. We want to know whether the mean of one group differs from the mean of another group, or whether two variables are correlated, or if a parameter in a regression model differs from zero.This article explains and discusses the idea of a bootstrap percentile confidence interval-which is an alternative to many of these methods. Furthermore, it is an alternative with substantial advantages over the usual methods. These advantages fall into two separate categories.The first stems from the use of confidence intervals in place of significance levels (p values) and null hypotheses. Instead of, for example, testing the hypothesis that there is no difference between the means of two populations, and citing the resulting significance level, we can set up an interval estimate for the difference of the two means. This has a number of advantages, which are briefly explained in the next section.The second category of advantages are those due to using the method of bootstrapping in place of conventional methods to derive confidence intervals. This is the main subject of this article. This second category of advantages divides into three subcategories:

Soil microbial community responses to the fungal endophyte Neotyphodium in Italian ryegrass

“…Each confidence interval was estimated by bootstrap with 4000 replacing re-samples (Wood 2005) using Infostat software (InfoStat 2008). We tested whether net catabolic response of soil differed significantly from zero.…”

When Male Norms Don’t Fit: Do Traditional Attitudes of Female Colleagues Challenge Men in Non-traditional Occupations?

“…Additionally, the rationale behind bootstrapped confidence intervals is very transparent, therefore sophisticated statistical concepts (e.g. normal-distributed standard errors as a condition of valid regression analysis) lose their priority (Wood 2005). Due to high interrelations between the three subscales of the MRNS (anti-femininity, status, and toughness norm) in each of the bootstrapping regression models a different attitude of female colleagues toward the male role was entered as a predictor with each psychological health outcome (depressive moods and job satisfaction respectively).…”

Fair Statistical Communication in HCI

“…Bootstrapping is a very useful method that has not received enough attention (Kirby and Gerlanc, 2013;Wood, 2004Wood, , 2005. Briefly, it consists of generating many alternative datasets from the experimental data by randomly drawing observations with replacement.…”