“Statistical Models and Shoe Leather,” David Freedman, 1991

A rebuke of statistical investigation taking the place of well thought-out research. Notes that, when applying basic statistics to data, it is almost never the case that all of the necessary assumptions of a process hold (i.e., linearity of the data in OLS), that without special care, causality is impossible to determine, and that for many research questions, a more detailed data collection process – particularly a hunt for natural experiments – will lead to more robust results than statistical observation alone. Provides examples from top polisci and sociology journals pointing out clear misuses of statistics.

http://www.rochester.edu/College/PSC/clarke/204/Freedman91.pdf

Advertisements
%d bloggers like this: