Trapiche Malbec 2006

on Oct 28, 2007
It felt almost as if I were drinking watered down blackberry juice steeped in an oak barrel. Yuk!

Bayes Theorem and Alzheimer's

on Oct 17, 2007
Apparently there is a new test for detecting Alzheimer's in its early stages:
One of the most distressing aspects of Alzheimer's disease is the difficulty in determining whether mild memory problems are the beginning of an inevitable mental decline. Researchers at the Stanford University School of Medicine have developed a blood test that is a step toward giving people an answer two to six years in advance of the onset of the disease. The test identifies changes in a handful of proteins in blood plasma that cells use to convey messages to one another. The research team discovered a connection between shifts in the cells' dialog and the changes in the brain accompanying Alzheimer's. They found that the blood test could indicate who had Alzheimer's with 90 percent agreement with clinical diagnoses, and could predict the onset of Alzheimer's two to six years before symptoms appeared. "Just as a psychiatrist can conclude a lot of things by listening to the words of a patient, so by 'listening' to different proteins we are measuring whether something is going wrong in the cells," said Tony Wyss-Coray, PhD, associate professor of neurology and neurological sciences and senior author of the study. "It's not that the cells are using new words when something goes wrong," said Wyss-Coray. "It's just that some words are much stronger and some are much weaker; the chatter has a different tone." [ScienceDaily]
The test is 91% accurate More than 10 percent of all people over 65 have Alzheimer's disease. I'll simplify it to 10%. If you test 1,000 people for Alzheimer's, 100 of them will develop Alzheimer's. Of the 900 who don't have Alzheimer's 819 are gonna be told correctly that they are not gonna develop the disease, but 81 of them will incorrectly be identified as likely to develop Alzheimer's. Of the 100 who will develop Alzheimer's 91 will be told correctly that they will develop the disease and 9 will be told incorrectly that they won't develop the disease. Of the 1,000 who took the test 81 + 91=172 had positive results, but only 91 out of those 172 will develop Alzheimer's So if the test result comes up positive, you have a 91/172 = 52% chance of developing Alzheimer's