This article is based on the 165th Cutter Lecture on Preventive Medicine, presented by Sir David Cox on May 3, 2017 at the Harvard T.H. Chan School of Public Health, Boston, MA, USA. Since 1912, the Cutter Lecture is delivered at Harvard's Department of Epidemiology, according to the bequest of John Clarence Cutter, MD (1851–1909).
I greatly appreciate the invitation to give this lecture with its century long history. The title is a warning that the lecture is rather discursive and not highly focused and technical. The theme is simple. That statistical thinking provides a unifying set of general ideas and specific methods relevant whenever appreciable natural variation is present. To be most fruitful these ideas should merge seamlessly with subject-matter considerations. By contrast, there is sometimes a temptation to regard formal statistical analysis as a ritual to be added after the serious work has been done, a ritual to satisfy convention, referees, and regulatory agencies. I want implicitly to refute that idea.
Bradford Hill A. The environment and disease. Proc R Soc Med. 1965;58:295–300.
Cochran WG. The planning of observational studies of human populations. J R Stat Soc A. 1965;128:234–65. CrossRef
Cornfield J, Haenszel W, Hammond EC, Lilienfeld AM, Shimkin MB, Wynder EL. Smoking and lung cancer. J Nat Cancer Inst. 1959;22:173–203. PubMed
Cox DR. Big data and precision. Biometrika. 2015;102:712–6. CrossRef
Cox DR, Battey H. Large numbers of explanatory variables: an exploration. Proc Natl Acad Sci (to appear).
Cox DR, Wermuth N. Multivariate dependencies. London: Chapman and Hall; 1996.
Fisher RA. The design of experiments. Edinburgh: Oliver and Boyd; 1935.
Student. The Lanarkshire milk experiment. Biometrika. 1931;23:398–406. CrossRef
Tibshirani R. Regression shrinkage and selection via the lasso. J R Stat Soc B. 1996;58:267–88.
- Statistical science: a grammar for research
David. R. Cox
- Springer Netherlands