Inference Factory

It’s interesting how the Phyicist E. T. Jaynes approached applied statistics. When faced with a new problem, he didn’t look up some standard solution, rather he derived the equations from scratch using the sum/product rules including approximations when needed.

Continue reading →

Researcher’s degrees of freedom

This is the fable of two statisticians on the battlefield. Two statisticians are on the battlefield and bullets are flying. One is a Frequentist and one is a Bayesian. Eventually, they both take one in the gut and blood starts gushing everywhere. A medic runs up to the wounded statisticians and says “you’ve both been […]

Continue reading →

The unshakable, yet completely wrong, myth at the heart of Statistics.

Like the supposed lack of atheists in foxholes, most statisticians believe everyone is a Frequentist when betting at a roulette wheel. No matter how sophisticated the Bayesian sophistry when you’re standing in front of the roulette wheel about to loose your life savings, your probabilities should at a least approximate the frequency at which each […]

Continue reading →

Preregistration.

It’s impossible to discuss preregistration with statisticians since their statistical indoctrination tells them to conflate two different things: (A) Policy X will improve the percentage of published claims which are true. (B) Policy X will improve individual inferences.

Continue reading →

Causal Research in Statistics is lame

I never mention the very popular causal research of either the Donald Rubin or Judea Pearl varieties on this blog. While thinking about causes is perhaps the most useful thing you can do in science, causal research of a certain kind found in statistics is kind of lame. On this score I share a prejudice […]

Continue reading →

The meteorologist’s job

The job of the meteorologist according to some is to produce forecasts such that “it rains X% of the time when they say there’s a X% chance of rain”. This view is held so strongly by Frequentists and almost all “Bayesians” that I doubt it will ever be eradicated. Frequentists in particular will go to […]

Continue reading →

The Logic of Science

In the title to his book, Jaynes claimed probability theory was the “logic” of science. This can be checked in hundreds of ways and it’s usually fun and insightful to do so. One of these was brought up in the post Why science necessarily involves a logical fallacy by Lars Syll. The idea is that […]

Continue reading →

Rediscovering how to do real science

In the previous post I discussed the latest lame brain idea for fixing science called “preregistration”. It was pointed out that preregistration may be correlated with “good science” but it’s not a cause of it. If you do science right you can throw preregistration in the same trash can you put p-value laden social science […]

Continue reading →

A modest proposal to fix science

Here’s a superb idea to fix science. First, gather together everything that correlates with good science. Maybe blond researchers publish worse papers. Perhaps if you work at a place beginning “University of” your papers reproduce less. Maybe 35 year olds publish papers that reproduce more than 45 year olds or 25 year olds. Whatever it […]

Continue reading →

Statistics is in a sad sorry state

If you want evidence of the staggering muddleheadedness of the statistics community, check out this exchange where three Ph.D.s struggle to understand estimating the speed of light from Michelson’s measurements. The bizarre thing is that while the they can get the estimate (well the philosopher probably can’t, but the others could) they’re not clear why […]

Continue reading →

Bayesians should leave Frequentists in the dust

The current debate between Bayesians and Frequentists involves Bayesians demonstrating real but limited improvements over traditional Frequentist methods. This can be a dangerous game for Bayesians to play as illustrated here and here. But even worse, this wastes Bayesian’s energies. The most powerful uses of statistics come from fully exploiting real Bayes in ways undreamed […]

Continue reading →

E. T. Jaynes: the book.

This is the first of a series of posts about the physicist E. T. Jaynes. His book on probability theory is available here and his collected papers here. While his book is widely read, his statistics papers get less attention. The series will mostly address the content of those neglected papers, but this first post […]

Continue reading →

Biggest mistake Bayesians can make II

This is the second post on the dangers of allowing Frequentists to transition to Bayesian statistics while retaining some of their old errors. The previous post is here. Probability theory done correctly is consistent with deductive logic and common sense. For instance if you use evidence twice it has no effect:

Continue reading →