More on Bayes

I think we are very familiar with probability being expressed in the ‘normal’ way: “the relative frequency of an event. If it occurs on average three out of four times, he or she will assign to it a probability of 3/4.” But to help understand the Bayesian way, think of probability as a way to give a belief a degree of certainty based on incomplete information; “all probability is conditional and subject to change when more data emerges.”

Once new information is available, we can use that prior knowledge we already had plus this new information to derive an updated assessment of the probability a conditional event will occur.

Reading this article from the Irish Times, Bayes’s Rule plays an imperative role. Calculating the probability of an event is a matter of life and death when it came to the 1983 launch of space shuttle Challenger which ended up exploding and killing all 7 crew members. According to the article, a Bayesian analyst estimated a 1 in 35 chance of a major accident occurring, but NASA’s estimate was 1 in 100,000 no doubt giving a false sense of security.



One thought on “More on Bayes”

  1. Reading this article it seems a bit strange that the author gives a Bayesian value calculated in 1983 and applies it to a launch in 1986. The article also says the shuttle flew 25 missions, a successful number of which must have occurred between these two times. I mention this because I am curious whether the Bayesian values changes at all. Given that the shuttle continued to fly successfully, wouldn’t Bayesian statistics calculate the chance of disaster to lessen, since the events observed continued to be positive. It seems Bayesian methods are well suited for the calculation of these unique types of probabilities (there aren’t many space shuttle launches to get a good sample of events from), but the shortness of the article on the methods employed may be a bit misleading about its true value.

Comments are closed.