Talk on Visual Inference at the Social Science Data Lab

This Wednesday, October 18th, I will present “Visual Inference for the Social Sciences” at the MZES Social Science Data Lab.

This talk introduces a remedy to the criticism frequently voiced against data visualization and exploration: that it may give rise to an over-interpretation of random patterns. A way to overcome this problem is the realization that “visual discoveries” correspond to the implicit rejection of “null hypotheses”. The basic idea of visual inference is that graphical displays can be treated as “test statistics” and compared to a reference distribution of plots under the assumption of the null. Visual inference helps us answer the question “Is what we see really there?” By so doing, it seeks to overcome long-standing reservations against visualization as merely “informal” approach to data analysis and the fear that beautiful pictures may in fact not correspond to any meaningful patterns of substantive scientific interest. The talk illustrates the application and benefits of this visual method by drawing on examples from the social sciences. A little lab exercise will encourage participants to try out visual inference in practice using the statistical programming language R.

When? Between 15:30 and (max) 17:00

Where? Room A-231.