Statistics and SIDS

I have just finished reading Statistics: A Very Short Introduction by David J. Hand.  Overall, there wasn’t much practical information to work with, but it did provide a very basic background in general statistical theory, Bayesian networks, and hidden Markov models.  Throughout the book, Hand emphasizes the success of modern computing in the field, gives practical uses for each concept introduced, and explores a range of real-world applications of statistics.

One of the real-world instances chosen by Hand is the probability a mother could lose two of her children to SIDS, aka crib/cot death.  In 1996, UK resident Sally Clark had lost her first son to SIDS.  When her second son passed away in the same manner in 1998, she was arrested and tried for murder. Pediatrician Sir Roy Meadow testified that the chance of two children from an affluent family suffering sudden infant death syndrome was 1 in 73 million. The reasoning for such a low probability was that the second instance of crib death was considered to be independent of the first. If the two incidents are independent, the probability of one instance (1 in 8500) can be multiplied by itself to yield Meadow’s figure. However, as an article written in Plus Magazine explains, it is more reasonable to calculate the conditional probability using Bayes theorem to determine the probability of her innocence.  Which they conclude to be 2/3, given number of births and double murders each year in England and Wales.  Unfortunately, it wasn’t until the second appeal that Sally Clark was exonerated of her charges and released from prison in 2003 after serving 3 years of her sentence.

It was with great frustration that mere hours after reading about Sally Clark’s ordeal that I heard a similar story play out on an NPR broadcast.  In 2008, 17-year-old Nga Truong was interrogated and coerced into confessing to the murder of her own child.  This time is was her brother that had passed alway when she was 8 due to SIDS.  Under the same presumption that two incidents of SIDS in one family is improbable, the detectives aggressively interrogated Nga Troung while accusing her of both deaths.  Troung had been locked up for 2 years before her confession was suppressed on the basis of coercion.  The 13 minute story and police recording is available through NPR.

Taking in both these incidents in the same day was very troubling, primarily because the police in Troung’s case were more determined to get a confession than get the truth.  Even more troubling is the fact that Sally Clark died of alcohol poisoning in 2007, having never recovered from the incident.  While people like Donna Anthony, Angela Cannings, and Trupti Patel were all found innocent after further review of their cases, one has to wonder how many mothers have to face this turmoil without proper defense.

Gallery Update

I updated the gallery plugin to use WP Photo Album.  I initially used a gallery plugin that pulled all my pictures from Picasa, but that became intolerable.   Despite having my photos automatically uploaded from my phone, the usability of Picasa through Google+ is very immature.  Also there’s no official Picasa application for Linux, and all my photos get resized upon uploading.  So considering Picasa caps free storage at 1Gb, prevents access to full-resolution images, and links everything back to my Google identity, I’ll just host all my pictures through my web host.

Now I only need to manually extract all my photos from my phone since I have no automated way of getting the full resolution copies on the web.  When I do, I will update the gallery with my latest photos of the Shenandoah National Park.  They’re pretty much all like the photo above, but with a much higher resolution.

Gödel, Escher, Bach Study Group

Although I have never been much of a fan of Hofstadter, his work has a loyal and faithful academic following. To be honest I was never very tolerant of his fans gushing about how their minds were blown by his revelation inspiring work. However, I think it’s time now that I put that aside and see what GEB is actually worth.

Reddit.com/r/GEB is doing a group read through starting January 17. As the ‘community’ posts related materials as well as self-serving fodder to garner ‘karma’, it may be a good opportunity to ride the wave of peer pressure to complete the novel this decade.

As it has been pointed out on Reddit, MIT’s OpenCourseWare offers video lectures on GEB, albeit aimed at a high school audience. So it might be a good time to dust off my copy and try to knock this one out before the next New Year.

P.S. It’s also worth noting that anyone who intends on following the subreddit for content should be aware the signal to noise ratio tends to be low, especially if the group grows substantially. During the Stanford AI class, the professors would create posts to communicate with the students directly, but it was hardly worth subscribing to due to the deluge of inane posts.