Gaining confidence with intervals: practical guidelines, advices and tricks of the trade to face real-life situations.

Abstract

Confidence intervals and measures of effect size are gradually becoming the standard way of reporting the results of statistical analyses in research articles, used instead of or in addition to p values. However, this shift in research practices barely affected teaching practices up to now. This paper is the third of a series written to serve as a general reference on the use of confidence intervals in quantitative social sciences. Its purpose is to provide guidelines, advices and useful tricks of the trade that will allow readers (a) to face most of the statistical problems emerging in real-life research settings and (b) to improve their understanding of confidence intervals and answer more efficiently their questions of interest. The first part of the article briefly introduces the basic elements of an approach based on confidence intervals: Calculations, interpretation, and hypothesis testing. The second part is an attempt to present some of the most important (but sometimes neglected) advanced issues concerning confidence intervals: Graphic representations, complex distributions, national surveys, the larger family of interval statistics (e.g., prediction intervals), and the Bayesian approach to probabilities.

Keywords

• Confidence intervals
• interval statistics
• guidelines
• graphic representation
• national surveys
• Bayesian approach