# Bayesian statistics theorem holds its own - but use with caution

##### June 7, 2013 by Lin Edwards report

(Phys.org) —In a Perspective in Science magazine this week, a Stanford Professor of Statistics re-examines Bayes' Theorem, its varying fortunes over the two-and-a-half centuries since it was proposed, and its current boom in popularity and likely future.

Bayes' Theorem was proposed by Thomas Bayes in the 18th century, and it combines newly acquired data with prior data to predict an outcome. In his paper, Professor Bradley Efron of Stanford University, presents the example of predicting whether twins are likely to be fraternal or identical in his overview of the .

In Professor Brad Efron's example, there are two categories of data to be considered: the newly acquired data (that sonograms show a pregnant woman is carrying twin boys), and the prior data (the fact that one-third of twins are identical). Identical twins are twice as likely to produce twin boy sonograms because identical twins are always the same sex while fraternal twins have only a 50:50 chance of being the same sex.

Bayes' theorem combines these data in the formula.

P(A|B) = [P(B|A) x P(A)] / P(B)

where P(A|B) is the conditional probability of A given B, and P(B|A) is the conditional probability of B given A. In the example, A is the twins being identical, a prior with probability = 1/3. B is the "sonogram shows twin boys". Genetics imply P(B|A) = 1/2, giving P(A|B) = (1/2) x (1/3) / (1/3) = 1/2.

The formula correctly predicts the twins have an equal probability of being fraternal or identical.

The theorem has proved its worth, such as in 2012 when it was used to successfully predict the outcome of the U.S. presidential election in all 50 states before the final vote counts were available. Despite its success it has always been regarded with some suspicion by , particularly because it has been used when genuine prior data is unavailable or uncertain.

Efron also compares more recent statistical theories such as frequentism to Bayes' theorem, and looks at the newly proposed fusion of Bayes' and frequentist ideas in Empirical Bayes. Frequentism has dominated for a century and does not use prior information, considering future behavior instead.

Statistical theorems are important because they are widely used in areas such as medical research. Efron reports that in his work as an editor of a statistics journal, he found around 25 percent of papers used Bayers' theorem and most were based on uninformative priors, which counts against it, but on the other hand he notes that the current environment of data being produced in "fire hose" quantities means Bayers' theorem could effectively connect disparate inferences.

Efron uses another example to explain the problem: a study of 52 men with prostate cancer and 50 healthy controls looked at the activity of 6033 genes in the hope of identifying genes expressed differently in the patients. They calculated a test statistic (z) for each gene, with a normal bell-shaped distribution if there was no difference between patients and controls, but with larger values if there were differences.

The resulting histogram looked normal except for 28 of the genes on the right of the distribution, having z>3.40 values. These could represent real or false discoveries, since some z values are bound to be large even in the null hypothesis (no difference between patient/control gene expression). The frequentist theorem predicts false discovery rate (FDR) to be below 10%, which means only 2.8 of these are false. The Bayes theorem suggests the probability of nullness is also 10%, but no prior evidence is used, and the prior is estimated from the data itself. Efron says this "statistical jujitsu" is Empirical Bayes, which is essentially a fusion of frequentist and Bayes ideas that says that when there are large numbers of parallels (as in the 6033 gene study), the data "carry within them their own prior distribution."

Efron warns that Bayes' theorem can be used if genuine prior information is available but caution is needed if there are uninformative priors. For parallel cases, Empirical Bayes methods can be used effectively.

More information: Bayes' Theorem in the 21st Century, Science 7 June 2013: Vol. 340 no. 6137 pp. 1177-1178 DOI: 10.1126/science.1236536

Abstract
The term "controversial theorem" sounds like an oxymoron, but Bayes' theorem has played this part for two-and-a-half centuries. Twice it has soared to scientific celebrity, twice it has crashed, and it is currently enjoying another boom. The theorem itself is a landmark of logical reasoning and the first serious triumph of statistical inference, yet is still treated with suspicion by most statisticians. There are reasons to believe in the staying power of its current popularity, but also some signs of trouble ahead.

## Related Stories

#### When errors improve performance: Model describes how experiences influence our perception

November 28, 2011

During estimation processes we unconsciously make use of recent experiences. Scientists from Ludwig-Maximilians-Universität (LMU) in Munich and the Bernstein Center Munich asked test subjects to estimate distances in ...

#### Researcher finds solution to problem in 19th-century theory of meromorphic functions

May 28, 2013

General theory of meromorphic functions in the complex plane began in the nineteenth century, when E. Picard proved his famous 'Picard's little theorem'. Then, in the 1920s, R. Nevanlinna created the modern theory of meromorphic ...

#### Fainting: All in the family?

August 6, 2012

Fainting has a strong genetic predisposition, according to new research published in the August 7, 2012, print issue of Neurology, the medical journal of the American Academy of Neurology. Fainting, also called vasovagal ...

#### Eighteenth century minister's mathematical theorem helps scientists map out cell signalling pathways

March 30, 2010

(PhysOrg.com) -- Researchers at the University of Glasgow have successfully exploited statistical inference and mathematical modelling to discover how cells communicate.

#### Study shows our epigenetics change rapidly in first year of life

May 27, 2013

(Medical Xpress)—The way our genes are activated changes rapidly when we are infants, a new study of twins at the Murdoch Childrens Research Institute has found.

#### Google pays tribute to 'Fermat's Last Theorem'

August 17, 2011

Google paid tribute on Wednesday to 17th century French mathematician Pierre de Fermat, transforming its celebrated homepage logo into a blackboard featuring "Fermat's Last Theorem."

## Recommended for you

#### Report proposes standards for sharing data and code used in computational studies

December 8, 2016

Reporting new research results involves detailed descriptions of methods and materials used in an experiment. But when a study uses computers to analyze data, create models or simulate things that can't be tested in a lab, ...

#### Amber specimen offers rare glimpse of feathered dinosaur tail

December 8, 2016

Researchers have discovered a dinosaur tail complete with its feathers trapped in a piece of amber. The finding reported in Current Biology on December 8 helps to fill in details of the dinosaurs' feather structure and evolution, ...

#### Fossilized evidence of a tumor in a 255-million-year-old mammal forerunner

December 8, 2016

When paleontologists at the University of Washington cut into the fossilized jaw of a distant mammal relative, they got more than they bargained for—more teeth, to be specific.

#### Archeologist claims to have found proof that Hebrew was the first written alphabet

December 8, 2016

(Phys.org)—Douglas Petrovich, an archaeologist with Ontario's Wilfrid-Laurier University in Canada has sparked controversy in the ancient history scholarly community by making claims that he has found proof that Hebrew ...

#### New study finds mammals during age of dinosaurs packed a powerful bite

December 8, 2016

Move over, hyenas and saber-toothed cats; there's a mammal with an even stronger bite. A new study by Burke Museum and University of Washington paleontologists describes an early marsupial relative called Didelphodon vorax ...

#### Scheduling leisure activities makes them less fun: study

December 8, 2016

Nothing ruins a potentially fun event like putting it on your calendar. In a series of studies, researchers found that scheduling a leisure activity like seeing a movie or taking a coffee break led people to anticipate less ...

##### Fabio P_
3 / 5 (2) Jun 07, 2013
Now, if only Bayesian statistics were taught to students since undergrads, rather than indoctrinating them to worship at the altar of p < .05. If only.
##### patnclaire
not rated yet Jun 07, 2013
The author writes, "The theorem has proved its worth, such as in 2012 when it was used to successfully predict the outcome of the U.S. presidential election in all 50 states before the final vote counts were available." I have never read that. Could the author reply and say where she/he got that information? It would contribute to the article.

##### Doug_Huffman
2 / 5 (4) Jun 07, 2013
Some of the resistance to Bayesian-ism is that it is the razor's bleeding edge separating frequentist statistics from the statistics of unique events.
##### DonGateley
1 / 5 (3) Jun 08, 2013
For some excellent Bayesian controversy see:

http://plover.net...yes.html
##### srikkanth_kn
1 / 5 (2) Jun 11, 2013
Interesting as to why a theorem - (i.e a mathematical statement) has to lead to a 'controversy' - (contradicting arguments).

Bayesian theorem puts in mathematical form of what is obviously logical - eg. two boxes A and B contain black and white balls in 6:4 and 3:7 ratio respectively, given that box A is chosen (event B), probability of getting a black ball (event A)is (obviously) 6/(6+4) but more formally [(6/(6+3)) *((6+3)/(6+4+3+7)) ]/(1/2). .

Repeat this in unbiased manner and you will definitely converge to 0.6 after repeated number of times. - I dont see any reason for controversy if we have well understood before applying in practice.