This paper is from Session 7G: Statistics for non-quantitative majors
which comes under Topic 7: Statistics education and the wider society
(Tuesday 13th, 14:00-16:00)
Luring non-quantitative majors into advanced statistical reasoning (and luring statistics educators into real statistics)
Presenter
Co-authors
Abstract
Introductory courses in statistics often progress from simple analysis to anova and interaction. The epistemology is close to elementary physics–simplify, decompose, analyse, then add the bits together again, and you will understand the phenomena. Most interesting problems in social sciences are multivariate, and variables interact in complex ways. Biological systems provide a better analogy–below a certain level of complexity, you destroy the problem by decomposition. The big statistical ideas in social policy are things like the shape of multidimensional data surfaces, effect sizes, interactions, limiting factors, extrapolation, and confidence intervals. Statistical significance is usually irrelevant; assumptions about distributions are risky. Non-quantitative majors are quite right to be wary of standard statistical models. We offer a plan to lure nonquantitative students into statistical reasoning by proving rich, multivariate data from large scale studies, along with media accounts (which are often simply wrong). Some ‘proof of concept’ will be provided.
