Data visualizations to support statistical inference

Mittwoch, 13. September 2017
17.00 – 18.30 Uhr

A 703

Fachbereich Psychologie

Prof. Dr. Gary McClelland (University of Colorado, Boulder)

In contrast to an explosion of modern and insightful methods for visualizing big and highdimensionaldatasets, the graphics presented in scientific journals are mired in historicaltradition. Publication graphics typically have almost no relationship to the statistical analysisthey purportedly represent. We begin by examining the “golden age” of data graphics beforewe had computers, then consider the uninformative statistical graphs we create with thepowerful computers we all have on our desks or in our laps, and finally develop new ideas forgraphing both between-participant and within-participant statistical analyses. We concludewith an example suggesting that many studies that have failed to replicate might not havebeen published had the reviewers seen better graphics. The presentation is almost entirelygraphical and accessible to a wide audience.