tag:blogger.com,1999:blog-7893272060787897238.post6618441348751724994..comments2024-03-19T08:42:45.690-04:00Comments on The Delaware Libertarian: Sottomayor. Ricci v DeStefano, and the curious difficulty involved in parsing today's newsSteven H. Newtonhttp://www.blogger.com/profile/09097470960863103473noreply@blogger.comBlogger4125tag:blogger.com,1999:blog-7893272060787897238.post-76030667520212183502009-05-28T21:44:20.388-04:002009-05-28T21:44:20.388-04:00If you want uniformity, try throwing dice.
Even t...If you want uniformity, try throwing dice. <br />Even there you will find a difference between theoretical and actual probabilities. <br /><br />And yet someone will say the dice are loaded, or better yet, racially motivated.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-7893272060787897238.post-29664877337547571032009-05-28T00:38:16.807-04:002009-05-28T00:38:16.807-04:00Miko's analysis is flawed. in particular, the firs...Miko's analysis is flawed. in particular, the first part where he assumes uniform scores for each class of participants does more to obfuscate than illustrate, because such an outcome was clearly impossible according to the stated facts so any calculations drawn from it are irrelevant. Further, an assumption that the passing scores were uniformly distributed w/o regard to race does not seem warranted given the much lower passage rate for blacks.<br /><br />The statement <I>"When New Haven officials analyzed the written test results, they found that the pass rate for black candidates was approximately half the pass rate of white candidates."</I> proves that at least 1 and no greater than half of the black firefighters passed each written test (otherwise the passage rate for white candidates would have been either 0% or greater than 100%). This means that 1-4 blacks and 2-33 white/hispanic candidates passed the captain test and 1-9 black and 2-58 white/hispanic candidates passed the lieutenant test. <br /><br />The Cornell article also mentioned a 70% cutoff for passing scores, but beyond that there is not enough information present to analyze further w/o making unjustified assumptions.tomhttps://www.blogger.com/profile/06653459162258850269noreply@blogger.comtag:blogger.com,1999:blog-7893272060787897238.post-3955291814564121312009-05-27T20:59:43.497-04:002009-05-27T20:59:43.497-04:00Actually, for reasons I am just too plain tired to...Actually, for reasons I am just too plain tired to elucidate tonight, your analysis does not work (although I do have to check and see whether I left out one key element from the Cornell analysis). But i'll post it tomorrow.Steven H. Newtonhttps://www.blogger.com/profile/09097470960863103473noreply@blogger.comtag:blogger.com,1999:blog-7893272060787897238.post-34818962800476297452009-05-27T18:46:05.026-04:002009-05-27T18:46:05.026-04:00I will leave it as an exercise for the student to ...<I>I will leave it as an exercise for the student to crunch the numbers and find out that in order for the highest African-American candidate to have placed 13th for Lieutenant and 15th for Captain (with the "rule of three" allowing the Board to promote only the highest scoring three candidates), means that these candidates also had to have performed below the norm on the oral interviews as well.</I>This is incorrect, from the information you've given. For simplicity of numbers, suppose that every white/Hispanic tester scored exactly 50% on each part, for 50% overall and that every black tester scored 25% on the written and 75% on the oral, for a 45% overall, thus putting their average lower despite their having done better on the oral portion.<br /><br /><I>Of the 41 applicants who took the captain exam, eight were black; of the 77 who took the lieutenant exam, 19 were black.</I>From this, the probability (assuming uniformity) of a black in the top 3 are 48.8% and 57.8% respectively. Assuming the results are independent, the chances that a black would be in the top 3 in at least one of the two categories is 78.4%, which is suggestive of bias if the assumption of uniform probability was correct (but by no means conclusive). However, there are potentially some reasons unrelated to the test itself why uniformity may not be the case (e.g., socioeconomic factors leading to different use of preparation materials, etc.).Mikonoreply@blogger.com