Good class today – lots of people came, and paid attention. We worked on false positives, starting with a hypothetical aids test with a 98% true positive rate and only a 5% false positive rate. To build the contingency table for a population of 10,000 people (better that than dealing with percentages) we need the actual incidence rate. Several students quickly found that on the web: globally 35 million infected.Since we knew the population was 7 billion we cold get the rate of 5 per 1000 without a calculator. Then we built the table and just about everyone was genuinely surprised and appalled that only one person in 9 who tested positive actually had the disease!
Then we split up to do homework exercises. There’s a really hard one on the reasons NOT to test prenatally for Trisomy 18 disorder even if you do test for Downs syndrome. In the former case the incidence is so small that the number of abortions of healthy fetuses that might follow on a positive test result seems outrageous. But the passage leading to this conclusion was so dense that the students needed lots of coaching to get there. That highlights a problem in writing the text. If we spell out the shape of each exercise the students will be more likely to be able to do it now but less likely when they encounter it on their own in N years. I need to write about this in the instructor’s manual.
blog home page