Somewhere at some point someone (hahaha!) suggested that sampling and polling and the like are similar to a cook taking a spoonful of their well-stirred soup to determine how it tastes. Assuming cooks do this and it works it seems just as amazing as deriving large conclusions from "such a small sample".Anyway, probably does not make any sense or have any relation to reality, but it was fun coming up with anyway!!
Well, to continue with and extend this metaphor to the study and Iraq. Imagine Iraq as a large pot of soup at a rolling boil. Now, you decide to take a couple of samples from various locations within the pot; not necessarily just with a spoon from the top but say you drop in your thimble-on-a-string and pull it up and taste it.
Now say that you get five samples this way. One tastes almost passable, three taste really bad and the fifth one tastes truly beyond horrible. What is the "taste mean"? Taste standard deviation? What do you think the soup tastes like "overall"? What about the "truly beyond horrible" sample? Is it an outlier? Do we throw it out? Do we have a large confidence interval? If so and we add in the outlier, does that increase our confidence interval? So much so, that maybe large portions of the soup actually taste quite yummy!?!
[ a fully caused & embodied blog ] [ Good Sense Without God ]
It is in the prosecution of some single object, and in striving to reach its accomplishment by the combined application of his moral and physical energies, that the true happiness of man, in his full vigour and development, consists. Possession, it is true, crowns exertion with repose; but it is only in the illusions of fancy that it has power to charm our eyes. If we consider the position of man in the universe,—if we remember the constant tendency of his energies towards some definite activity, and recognize the influence of surrounding nature, which is ever provoking him to exertion, we shall be ready to acknowledge that repose and possession do not indeed exist but in imagination. - Wilhelm von Humboldt, The Sphere and Duties of Government (The Limits of State Action) (1854 ed.)
Friday, July 27, 2007
Sampling Soup: Iraq and the Lancet Study
In regard to this debate of the study by Roberts el al. in 2004, I will weigh in with a modification of an analogy I ran into somewhere. By the way, I have ZERO expertise in statistics soooo ... fortunately this is my blog and I can play one if I wanna! grin!