"Real men keep p values to themselves"

A short lesson in statistics


Andrew Leonard
January 15, 2009 12:28AM (UTC)

In "The Amazing Shrinking Citigroup," posted Tuesday, I asked my readers to explain Felix Salmon's statement that "I'd say p=0.3 right now that Barack Obama's first major act as POTUS will be the nationalization of Citigroup." I confused matters by transcribing "p=.3" as "p=.03," as a number of readers immediately pointed out.

Two other regular readers responded by private e-mail, and I'm sharing those messages with you because they were elucidating, and funny.

Advertisement:

P=.03 should mean that the probability of the event occurring is equal to the 97 percent confidence interval. Since most statistical analysis, ie that not done by methodologists, uses p < .05, i.e. the 95 percent confidence interval, as the cut off for significance, this appears to be a very specific call out of what should be considered a statistically significant event. Personally, I'd like to see the betas, bs, and standard errors so I could see just how much movement change in Y there is for every unit of change in X, as well as the directions.

Adam L. Silverman, PhD, Human Terrain System, US Army -- TRADOC

"The views expressed here are mine alone and do not necessarily reflect those of the US Army's Human Terrain System, the Training and Doctrine Command, and/or the US Army".

---------

P -- as in "p-value" or:

The probability of what you have observed in the data, if your model of the world is true. People generally want this value to be small: a small p-value implies that the reduced model is unlikely, that your observations suggest something non-null is afoot.

People pick a cutoff -- arbitrarily, really -- and a p-value lower than the cutoff is pronounced "Significant." Alchemy!

A large p-value, alternately, is uninteresting: the data is consistent with what you already know.

Frankly -- p-values are a bit passe, if not completely gauche -- statistically speaking. Modelling, these days, is more particular (if not exact). Macho statisticians are proud of tight (posterior (bayesian)) confidence intervals. Real men keep p-values to themselves.

Kobi Ako Abayomi, Asst. Professor Georgia Institute of Technology, School of Industrial and Systems Engineering -- Statistics Group


Andrew Leonard

Andrew Leonard is a staff writer at Salon. On Twitter, @koxinga21.

MORE FROM Andrew LeonardFOLLOW koxinga21LIKE Andrew Leonard


Related Topics ------------------------------------------

Globalization How The World Works

Fearless journalism
in your inbox every day

Sign up for our free newsletter

• • •