Comments Posted By Jurgorr
Displaying 1 To 1 Of 1 Comments

A MOST GHOULISH DEBATE

DrSteve:

I fully agree with you that the clustered sampling is not as good as a truly randomized sampling would be (in terms of tightest error bars for a given number of samples). But I fail to see why this methodology would either over- or under- estimate the death toll. If there is spatial correlation, it's true that you're more likely to find deaths close to other deaths. But you're equally likely to find survival close to other survival. Hence, there is no push one way or the other, from what I understand. There is probably a reason it's called "spatial correlation", not "spatial bias".

As an aside: It bothers me that people are willing to accept a concrete "lesser evil" like killing Iraqis over a completely imaginary "greater evil" like the WMD bogeyman, and that somehow this gets passed off as "logic". Doesn't matter how logical and rational you are: Garbage-in, garbage-out. Seems considering uncertainty in decision making is too abstract and difficult for some people's minds (especially those used to spoon-fed religious dogma).

Salvage: I loved your post 48.

Comment Posted By Jurgorr On 13.10.2006 @ 02:24


 


 


Pages (1) : [1]


«« Back To Stats Page