My paper on data visualization next week may be the most useful thing I do this year, if I succeed in convincing my audience, that is.
Fact: Statisticians have failed in a very important respect.
This fact became apparent to me on a beautiful day lying on the beach in Santa Monica when it was over 80 degrees in late December. Some very well-respected polls have shown that 26% of Americans believe global warming is false.
As statisticians, we spend much of our careers doing things “right”, as defined by our peers, that being other people who almost always have Ph.D.’s and frequently have supercilious attitudes. We’re concerned about accounting for stratification because that will provide the correct standard error which will in turn give us the right significance level. We need to consider fixed effects versus random effects, the normality of the distribution and so on to a great degree.
The fact – and we KNOW this – is often those factors we are using to tear up our colleagues’ work are not substantively important. Yes, they may have inflated their standard error by a factor of three but if it was .0003 instead of .0001 , who the hell really cares.
Even if everything was done perfectly, if your article reads like a combination of instructions for replacing my hard drive, the Odyssey in the original Greek and my federal schedule C attached to my 1040 tax return – then you failed.
We’ve failed at telling our story. Twenty pages of tables of numbers complete with coefficients, effect sizes, training sets, test sets, F-tests, t-tests, eta-squared and more may be perfectly well done and impress people with our brilliance.
Listen to me very closely my fellow statisticians…
The goal here is not to get people to admire you.
It is to get them to believe you.