Letter to the editor. NAA and JFK: Can revisionism take us home?

The Annals of Applie d Statistics 2008, V ol. 2, No. 1, 430–431 DOI: 10.1214 /07-A OAS153 c  Institute of Mathematical Statistics , 2 008 LETTER TO THE EDITOR NAA & JFK: CAN REVISIONISM T AKE US HOME? By John E. Fiorentino Occasionally during the course of the human learning exp erience we are faced with an a nomaly . An ab erratio n of sorts, whic h try as we might, d efies appropriate classification. The recen t paper b y Sp iegelman et al.— Chemic al and for ensic analysis of JFK assassinat ion bul let lots: Is a se c ond sho oter p ossible ?—is on e such ab erration. It is riddled w ith b oth m isconceptions and errors of fact. Purp orting to cast doubt on the NAA (neutr on activ ation analysis) work conducted by Dr. Vincent Guinn in the inv estigatio n of the assassination of President John F. Kennedy , it fails miserably . The pap er offers t w o central conclusions, one wh ic h is demonstrably false, and th e other whic h is sp ecious. The authors opin e; “ If the assassination fr agments ar e derive d fr om thr e e or mor e sep ar ate bul lets, then a se c ond assassin is likely, as the additional bul let would not b e attributable to the main su sp e ct, Mr. Oswald. ” This statemen t relating to the lik eliho o d of a second assassin based on the p remise of three or more sep arate b ullets is demonstrably false. The a v ailable evidence ind icates that Oswa ld fired three shots, one of wh ich is b eliev ed to ha ve m issed. Ho w ever, on the off chance that all three shots hit (ev en though there is absolutely no other supp orting forensic evidence for suc h a n otion) those thr ee shots alone in no wa y would in d icate then that “a second assassin is lik ely .” The authors’ err oneous conclusion was ac h iev ed b ecause th ey hav e either b een misled (which I p ersonally b eliev e is the case) or they simply aren’t familiar with the evidence. The second fatal fl a w is the u s e of a rather u ncomplicated formula based on Bay es Theorem. Through this form u la the authors conclude: “thus the critical ratio = .53 ov er .80. Sinc e this r atio is less than 1, Dr. Guinn ’s testimony that the evidenc e supp orts two and only two bul lets making up the five J FK fr agments is fundamental ly flawe d. ” Received Octob er 2007. This is an electro nic reprint o f the or iginal article published by the Institute of Mathematical Statis tics in The Annals of Applie d Statist ics , 2008, V ol. 2, No. 1, 43 0–431 . This re pr int differs from the or iginal in pag ination and typogr aphic deta il. 1 2 J. E. FIO RENTINO Sev eral problems immediately present th emselv es. The authors are using the metho d kno wn as Ba yes F acto r s; this is legitima te and th e equation they quote is deriv ed from Ba yes’ la w. The fir s t term is th e p osterior od ds of 2 bullets versus 3; the s econd term is w hat is kno wn as the “lik eliho o d ratio,” not “ratio of probabilities” as the authors claim. It mak es no sense to talk ab out probabilities of evidence; bu t th is is a matter of fu ndamenta ls n ot metho dology . The third term is the prior o d ds. In triguin gly , the p rior o d ds the authors men tion p ertain to multiple sho oters v ersu s a single sho oter. It should b e the pr ior o dds of 2 bu llets v ersus 3 or more, b ecause th e p osterior o dds p ertain to th ese ev ents. Ho w ever, nowhere in the p ap er do you see wh at prior o d ds were us ed . The ratio 53 to 80 is, from a Ba yesia n p oin t of view effectiv e only when it is mo d ulated by th e prior o dds. By using a misguided and biased approac h to this very imp ortant evi- dence, Spiegelman et al. h a ve effectiv ely negated their find ings. In a nutshell th e ONL Y evidence of an y other “bu llets” MUST come fr om the sev eral minute particles of lead reco v ered fr om th e victims and the crime scene, as we ha v e evidence of 2 bullets which are ballistically matc hed to th e Osw ald rifle to the exclusion of all other weapons in the wo r ld. So, IF there is another b ullet the ONL Y evidence of its p assing is a p article (or particles) of lea d. Unfortunately , for that idea, this “bullet” m ust ha ve hit someone or something in the limousin e. Unfortunate again is the fact ther e are no w oun ds in either victim to w h ic h it could b e attributed, nor any damage observ ed in the limousin e, to whic h it could b e attributed. S o, essent ially w e are left with a “bullet” wh ic h didn ’t h it either victim, didn’t hit anywhere in the limousine, and only left b ehind a p article (or particles) of lead as evidence of its passing. No w, I’m n ot exactly s u re of the “statistica l pr obabilit y” for that even t, b ut my feeling is the answe r lies somewh ere outside the domain of either statistics, or c hemical analyses. Perhaps it migh t come fr om the examination of tea lea v es, lines on the p alm of one’s hand, or b u mps on the head. But wherever yo u may find it, it wo n’t b e “science.” Fiorentino Research PO Box 3 24 Oakhurst, New J ersey 07755 USA E-mail: johnfioren tino@optonline.net URL: www.fiorentinoresearc h.com