UTCS Colloquia - Keith Winstein/MIT, "The non-conflict between Bayesians and Frequentists", ACES 2.402

Contact Name: 
Jenna Whitney
Apr 22, 2011 1:00pm - 2:15pm

There is a sign-up schedule for this event that can be found at



Type of Talk: UTCS Colloquia

Speaker/Affiliation: Keith Win


Talk Audience: UTCS Faculty, Graduate Students, Undergrads

, and Outside Interested Parties

Date/Time: April 22, 2011, 1:00 p.


Location: ACES 2.402

Host Name: Michael Walfish

Talk Titl

e: The non-conflict between Bayesians and Frequentists

Talk Abstract:

In 2007, an academic cardiologist downloaded 42 medical studies from th

e Web site of drug giant GlaxoSmithKline, combined them, and found that t

he world''s best-selling diabetes drug caused heart attacks. GSK lost about
$12 billion in sales and market value. But a different way to analyze the

same data, called Bayesian analysis, finds with conservative assumptions

that the drug actually reduces heart attacks. Or does it? I''ll present a n

ew take on the difference between classical statistical methods and Bayesia

n techniques, showing how the opposing schools are two sides of the same c

oin. Viewed formally in the context of confidence and credibility intervals

, criticisms of each practice have a tight symmetry and can be transformed
into each other. I''ll also show results from a new algorithm that calcula

tes the performance of contemporary "exact" hypothesis tests that hadn''t p

reviously been characterized.

Speaker Bio:
Keith Winstein covered s

cience, technology and medicine as a staff reporter for The Wall Street Jo

urnal from 2007 to 2010. He is now pursuing a Ph.D. in electrical engineeri

ng and computer science at MIT.
Keith created the Library Access to Music
Project, which has served free music on-demand at MIT since 2004. He is t

he co-author of the shortest DVD descrambler, qrpff, and of the analysis

that cracked the encryption on Motorola RFID cards. In 1998, Keith''s stud

ent paper initiated the field of lexical natural language steganography, o

r programmatically hiding secret messages in plain text.