Today in The New York Times Magazine Maxine Swann tells the curious story of Paul Frampton, a 68 year old theoretical particle physicist who was apparently duped into becoming a drug mule by a bikini model he met online. The story is a fascinating tale of a giant academic ego and the seemingly infinite gullibility of this scientist.
Something stood out in particular for me. During the trial, Frampton was asked about several notes and calculations that were found on him when he was arrested. He had jotted: “5 standard deviations 99.99994%”, which he explained in court to be the criterion for the discovery of the Higgs Boson; a result that is unlikely to occur due to chance. He further explained that he was “calculating the probability that Denise Milani would become my second wife, which was almost a certainty.” Apparently, he took the messages and love notes that he had exchanged online with the purported ‘Milani’ to be strong evidence that she loved him. Under the null hypothesis — she doesn’t love me — these behaviours would have been very unlikely indeed.
Aside from committing the p-value fallacy, what else is wrong with Frampton’s logic?
The fact that Frampton was being set up was immediately obvious to his friend, who warned him about what was up in no uncertain terms. Most of us would have taken all the information available to us to make a conclusion. How often do young bikini models fall for older professors with a poor relationship track record, for instance? However, Frampton choose to only use a select set of observations on which to make his inference. Had he have incorporated prior information, or updated his beliefs as new evidence became available, he may have been able to avoid his 5 sigma mistake, and the nearly 5 years of jail time which he was sentenced for it.
Quite incredible, not really sure what I just read: a physics professor arrested with a black back which was not his, despite several warnings and 15 (15!) days of chaos with airline tickets!
Pingback: P-value fallacy alive and well: Latest case in the Journal Nature | bayesianbiologist