6-825msg3.txt

运用贝叶斯网编的一个识别垃圾邮件的程序

Subject: re : 6 . 797 , comparative method : n - ary comparison

gotcha ! there be two separate fallacy in the argument against n-ary comparison which i discuss recently and which power , delancey , and guy be now apparently seek to defend . ( 1 ) janhunen say that the probability of a match occur purely by chance when you compare japanese with four language be four time what it be when you compare it with one language . this simply cannot be true because probability be value between 0 and 1 . if the probablity in the case of a binary comparison be say . 5 , then he would be predict that it would be 2 in the case of n-ary comparison , which be impossible , because 2 be not between 0 and 1 . ( 2 ) the other fallacy be not purely mathematical , although i suspect that it involve element of confusio . in any case , no one who argue for n-ary comparison ever talk about get a match in 2 out of n language . now , if we look at guy 's number , in his scenario of a 100 - word list with no shift meaning , he come up with 14 . 5 probable spurious mathce in a binary comparison but only 5 . 8 when you be look for a match between 3 out of 5 language , 0 . 13 when you look for one between 4 out of 5 , and he do not give the much smaller number yet in the case of 5 out of 5 . i be not sure how jacque define spurious and so i have not verify the number , but they be certainly on the right order of magnitude . as you consider more and more language ( also as the initial probability of a match decline , which usually happen as you go from toy model to real datum ) , what happen be that you need fewer and fewer out of the n language be compare to agree . thus , in guy 's example a match between n - 2 language out of 5 be less likely to occur by chance than one between 2 out of 2 . but if n be 100 , i . e . , you be compare 100 language , then you would not need n - 2 ( i . e . , 98 ) language to agree to be able to do better than with a binary comparison . it would be many many fewer ( although i do n't know how many since i do not know what formula jacque be use and what he be assume about the initial probability of a match ) . maybe , he could kindly supply the number . and in light of all this , let us add another argument for reject indo - european : bopp never offer a mathematical demonstration that the relationship he propose be unlikely to be due to chance , much less by do a binary comparison of every pair of indo - european language . which i think just go to show how unrealistic the whole idea of do such comparison be . but if you do want to do them , then at least let us be clear about how to do them so as to minimize false positive ( i . e . match due to chance and not really reflective of common origin ) as well as false negative ( i . e . , failure to find genuine historical connection ) . on the second point , there be argument that n-ary be better . alexis mr