Statistics and Epidemiology

In the 1930s English statistical theory was beginning to travel, with contributions from, amongst others, Hotelling and Snedecor in America and Darmois in France, but its home was still in England where there were four important centres: University College London, Rothamsted Experimental Station, Edinburgh University and Cambridge University with University College and Rothamsted far in the lead. Although Cambridge University was slow to adopt modern statistical theory, Cambridge men–Karl Pearson, Edmund Whittaker and Ronald Fisher–had put the other places on the statistical map. University College was the most established centre and its importance went back to 1893 when Karl Pearson, the professor of applied mathematics, first collaborated with Raphael Weldon, the professor of zoology on a subject they called “biometry.” There was a second surge in the “English statistical school” associated with R. A. Fisher who went to work at Rothamsted in 1919.

I wish that people would be persuaded that psychological experiments, especially those on the complex functions, are not improved (by large studies); the statistical method gives only mediocre results; some recent examples demonstrate that. The American authors, who love to do things big, often publish experiments that have been conducted on hundreds and thousands of people; they instinctively obey the prejudice that the persuasiveness of a work is proportional to the number of observations. This is only an illusion.

While it is easy to lie with statistics, it is even easier to lie without them.