Friday, March 03, 2006

Hurricanes & False Positives

So now we are treated to video tapes of President Bush and his advisors being warned that Katrina was, in the "gut" of FEMA's then-head, 'the big one.'

It's all very well and good to judge in hindsight. I was driving yesterday and heard a news report on a local liberal classical music station that New Orleans Mayor Ray Nagin 'had a sinking feeling' when he heard of, or watched, the videotape. Nice dramatic touch- it's clearly his style. Find a scapegoat and pile on, while hoping everyone will forget your own culpability.

In reality, though, the situation was not nearly so clear prior to Katrina's landfall.

Let's remind ourselves of the two risks which exist in every hurricane situation. They are familiar to any data analyst. I am speaking of Type 1 and Type 2 error.

Type 1 error is that against which statisticians are trained to guard. It is the risk of rejecting an hypothesis which cannot, in fact, be confidently rejected. Most statistical tools focus on the rejection of hypotheses at some "confidence level." In this case, it is the risk of rejecting the premise, "Katrina is the ultimate lethal hurricane," or something similar, when, in fact, the premise is found later to be true.

Type 2 error is the risk of accepting a false hypothesis.

If Katrina had been judged, before it made landfall, to be as deadly as it ultimately was, then the correct decision would have been to order massive evacuations. If it turned out to be rather small instead, then authorities would have made a Type 2 error. They would have accepted the premise, "Katrina is the ultimate lethal hurricane," when, in fact, it was not. Don't think that great ridicule and anger would not have been unleashed on all authorities, beginning with President Bush, in that event. Ray Nagin would have been screaming, '...they told me to evactuate. it's not my fault that people died, were injured, and property was looted in the rush out of town.....'

As it now stands, authorities believed they should wait until they were more sure of the hurricane's strength than, evidently, the date and time of the recently-released videotape. They minimized Type 2 error, accepting a false premise, in favor of not needlessly disrupting a major city for a non-lethal hurricane. Either way, the administration was going to be criticized.

It's convenient, in the aftermath, for the mayor of New Orleans, and anyone else with a liberal-leaning agenda, to declare that the Federal authorities "knew" all along how deadly Katrina would be. But that's just not true, and it's an oversimplification of what most people understand is an always-difficult decision- if and when to evacuate a large population center in advance of a hurricane, attempting to balance the probable costs in lives and property of evacuation with those of staying put.

No comments: