.. . . The last sentence demonstrates the worthlessness of his chosen method. The error range (97,808,000) is larger than his "estimate" (61,911,000). I don't have to be a statistician to know that when your estimate is smaller than your error range, your method is about as good as a ouija board . . . >> Rummel's use of the term "error range" is not the same as that implied by the writer. The writer, unbeknownst to himself, is referring to the standard deviation of a set of numbers. The range or gap between high and low estimates has no bearing on the reliability of any "point estimate" (i.e., a single number). The standard deviation does. Of course, whatever the standard deviation, the mean is going to be the most likely estimate, broadly and relatively speaking. (Been a few years since I taught stats.) I take it that Rummel's bag of estimates have no known probability distribution, so assuming they are all equally likely or unlikely, the only reasonable thing to do is take an average. But their likelihood is no better or worse than the historical account underlying the number(s). There is no statistical exercise, in the sense of some kind of inference from data based on statistical theory, to criticize here, since there is no data that could be used for such an exercise. I'll forego responding to the rest of your post. I know you won't miss it. I will note my impression that everyone is dismantled on alt.politics.socialism.trotsky at one point or another. I have a unique escape from these tedious debates on how many zillions were cruelly exterminated at the hands of fascism, communism, or imperialism. I simply disclaim support for any of them and try to live different. It's not that hard, actually. just my tinkle in the river of discourse. mbs
