On lessons learned, from Nate Silver
Jan. 26th, 2011 02:30 am![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
tahnanA few weeks ago, Nate Silver of 538 wrote about trying to draw lessons from the Giffords attack:
It's also something I've been thinking about the last week in terms of the Mystery Hunt, and yes, I know that "20 injured, six dead" to "Mystery Hunt" is the world's worst trivialization. But in thinking about things our team might have done wrong this year (in terms of, let me stress, maximizing our fun, not in terms of winning), I have to remind myself that it's essentially a single data point; if something went wrong this year that hasn't in years past, I need to not rush to "correct" it. Think about it, yes; consider whether there was a policy in place that failed, yes; assume the same thing will happen next year, no.
Sorry, that's your random thought for the day. And speaking of random thinks for the day, Jan Freeman of the Boston Globe posted in her blog the Google N-Gram result for "another thing coming" vs. "another think coming". Well, I liked it, anyway.
Suppose, however, that our focus is on something slightly different: on estimating the risk posed by assassinations of public officials in the future, and thereby doing our best to prevent them. Here, there is probably the risk of overgeneralizing the lessons learned from just one attack, however terrible it was.I thought that last sentence was remarkably insightful. I mean, Silver is a statistician; it's no surprise that he's had that insight about statistics, and I'm sure he's by no means the only person to observe it. (cf Steven Brust: "Everybody generalizes from one example. At least, I do.") But it did put the Tucson shooting into perspective for me. There are certainly things we can learn and policies we can thereby implement, policies on things like gun control and mental health care. But trying to draw conclusions about the nature of political violence from the attack is a mistake; it's much like, I think, the old adage about generals fighting the last war.
[...some discussion of the rarity of assassination in America, then...]
These assassinations and assassination attempts, including the one against Ms. Giffords, cannot be thought of as “random” incidents. By definition, an assassination requires that a specific public official be the target for murder, and the authorities say Mr. Loughner seems explicitly to have targeted Ms. Giffords.
They have, however, been isolated incidents in several different senses of the term: because of their rarity; because there is little to link them together; and because virtually all were the acts of lone individuals, and not part of larger conspiracies.
That makes them, for someone concerned with estimating the risk of similar future attacks and preventing them, exceptionally difficult to study. It is almost impossible to come to meaningful conclusions about probability from a sample size of one. [emphasis added]
It's also something I've been thinking about the last week in terms of the Mystery Hunt, and yes, I know that "20 injured, six dead" to "Mystery Hunt" is the world's worst trivialization. But in thinking about things our team might have done wrong this year (in terms of, let me stress, maximizing our fun, not in terms of winning), I have to remind myself that it's essentially a single data point; if something went wrong this year that hasn't in years past, I need to not rush to "correct" it. Think about it, yes; consider whether there was a policy in place that failed, yes; assume the same thing will happen next year, no.
Sorry, that's your random thought for the day. And speaking of random thinks for the day, Jan Freeman of the Boston Globe posted in her blog the Google N-Gram result for "another thing coming" vs. "another think coming". Well, I liked it, anyway.
