Andrew Gelman

http://andrewgelman.com/2013/04/24/the-tweets-votes-curve/

- Thread starter TheEcologist
- Start date
- Tags statistical quote

Andrew Gelman

http://andrewgelman.com/2013/04/24/the-tweets-votes-curve/

Chapter one (Probability Theory) quote: "You can, for example, never foretell what any one man will do, but you can say with precision what an average number will be up to. Individuals vary, but percentages remain constant. So says the statistician." (Sherlock Holmes in The Sign of Four)

I don't have the exact wording he used but Justice Scalia in rejecting the use of statistics in class action suits (a critical decision which gets little attention) argued that it had no validity in law because it was impossible to rule out alternative causes entirely with statistics.

While true, you wonder why the Supreme Court allows statistics to be used in trials at all (as it is) with that logic. Because by that logic statistics is pretty much useless.

valid defense of using non-Bayesian methods, namely incompetence

Inside every nonBayesian there is a Bayesian struggling to get out

"good Bayesian does better than a non-Bayesian but a bad Bayesian gets clobbered."

Some bayesians are like preachers, they take the credit to themselves and divert the blame to priors **like** preachers divert difficult questions to god

There is a trade off, outside academics, for improved results versus greater complexity. Most IMHO are going to chose the least complex method and are not going to use a methodology their boss will never have heard of in their one statistic course. I get lectures from my boss all the time for using overly complex methods or explanation and I don't even consider something like baysian which is never addressed in early statistics courses (actually I never encountered it even in a graduate program in measurement and statistics which included courses in SEM, ANOVA, HLM etc).

I have fallen far from the true faith of academics

I would even take away the word “Bayesian”.

As a former academic I think a better question is what percent of the people who use statistics for a living even know what [....] statistics is, let alone use it. I would suspect it is far less than one percent.

I always imagine your boss as a Dilbert-esq kind of character saying things like "We didn't hire you for your skill set! We hired you because I'm too lazy to run the incredible simple model myself!"

In 1930:ies R A Fishers boss wrote a book advocating deterministic allocation. The year after, Fisher wrote a book where he advocated randomisation – in contrast to his boss. It is not only Fishers brilliancy but also the qualities of his boss, that makes Fisher a famous person, and the rest of us not as famous.

The Dilbert bosses! (Choose your boss carefully!)

Our organization thinks statistical analysis is important - but senior management has limited personal interest in it [which is strange to me]. That is why I was hired (although obviously I am not a statistician).

So is nothing really like the Dilbert boss. It is just that when I discuss his, likely very rational take, on the organization it is always in the context of statistics here. Even among academics, in my observation, really bright people are commonly disinterested in expert knowledge in other areas.

Standing on the ground, looking up on the giants, who are standing on other giants shoulders, and I realize that they can see an ocean of knowledge, and I can only see a small pond, I would be most happy if I had just a percent of their knowledge.

"If I have not seen so far it is because I stood in giant's footsteps" - Linux Fortune

Which is inspiring, and a warning, sometimes focusing too much on what the great names of old have done, will blind you to what is obviously the best path forward now.

If you only follow great men, you will never leave their shadows.

Amusingly, or not, when Newton was the head of the Royal Society dealing with sciences he was a fierce reactionary who resisted new breakthroughs in science.

"...SES is sometimes said to be applicable for series showing no seasonal variation or long term trend but with a locally constant mean which shows some 'drift' over time whatever that means."