PICK A NUMBER — PICK **ANY** NUMBER
Posted by Henry Bauer on 2008/02/07
Statistics may seem to be plain information, but often they are issued in order to send messages. So the choice of numbers is naturally determined by what the desired message is.
Want to show that the HIV/AIDS epidemic is even worse than we thought?
“AIDS advocacy groups say the new figures [to be released by the Centers for Disease Control and Prevention] will put the number of Americans infected with the AIDS virus each year close to 50 percent higher than previous estimates, at 55,000 instead of 40,000”
[WASHINGTON (Reuters) “Under 1 percent of U.S. adults have HIV: report”, by Maggie Fox; 29 January 2008]
Want to show that the epidemic is not as bad as we thought?
“The CDC has estimated in the past that more than 1 million Americans in total are infected with the human immunodeficiency virus that causes AIDS.”
“In 1999 to 2006, the prevalence of HIV infection among adults aged 18-49 years in the civilian noninstitutionalized household population of the United States was 0.47 percent . . . . [which means] anywhere between 447,000 people and 841,000 people, with 618,000 the middle number. . . . . [according to] the National Center for Health Statistics . . . . The agency’s snapshot of HIV infection in the United States shows the rate continues to be stable. . . . The report covers adults aged 18 to 49 and only people living in households — not prisoners, the homeless or patients in institutions” [emphases added]
Want to show that there’s still ample cause for concern, even though the HIV/AIDS epidemic is not as bad as we thought?
Even if overall the numbers are stable—“We can say the prevalence is basically stable in this U.S., household-based population”—, it’s always possible to pick out sectors where there are grounds for grave concern:
“We do see the disparities by race/ethnicity . . . confirms other surveys that show black men are far more likely than other Americans to be infected. . . . Black men aged 40 to 49 had the highest rate of infection, at close to 4 percent” [emphasis added].
Also, “2% of non-Hispanic blacks were HIV-positive, compared with 0.23% of whites and 0.3% of Mexican-Americans” (Kaiser Health Disparities Report: A Weekly Look At Race, Ethnicity And Health, 31 January 2008).
Want to show that we must leave these important matters to the experts, because lay people can’t handle such complicated calculations?
Just try to put all the official numbers about HIV/AIDS together. The result will be great respect for the experts’ ability to remain unruffled in the face of blatant contradictions. For instance, the number of new infections is worse than we thought, more like 55,000 annually. Therefore the total number living with HIV/AIDS should have been rising at 55,000 per year. Yet we‘ve just had official assurance that the rate of HIV among Americans has been stable for something like a decade. What’s been happening to those 55,000 new cases annually?
A natural thought is that there have been 55,000 deaths from HIV disease per year. But if you look at the official statistics, you get the following (for 2004, from National Vital Statistics Report, 56 #5, 20 November 2007):
Deaths from HIV disease: 5608 white Americans, 7271 black, 84 Native American, 100 Asian Americans; the total of which is 13,063. (By the way: “HIV disease was also briefly among the top five killers for the black population during the 1990s” [emphasis added]).
Or by ethnicity: 1758 Hispanic deaths, 11,195 non-Hispanics: total is 12,953. (Again: “HIV disease was one of the top five causes of death for the Hispanic population in the mid-1990s but quickly dropped out of this group”.)
[13,063 minus 12,953 equals 110 people who are neither Hispanic nor non-Hispanic, apparently, but who died anyway.
The report does note, in very fine print, that reporting of ethnicity and race can be somewhat inconsistent, but annoying little inconsistencies like 110 apparently missing people could all be avoided if these reports didn’t claim more accuracy for the numbers than they deserve. Since uncertainties exist in the reporting of ethnicity and race, how about not giving numbers to 5 significant figures and just publishing “13,000”? That would be more honest as well as less annoying. Perhaps statisticians employed by federal agencies could take a short course in the use of “significant figures”—see MATHEMATICAL AND STATISTICAL LIES ABOUT HIV/AIDS, 2 December 2007.]
Anyway: About 13,000 a year die from HIV disease. About 55,000 newly contract it each year. Therefore the total number living with HIV/AIDS should be rising at the rate of about 42,000 per year. Yet the total number of Americans living with HIV/AIDS has remained stable since the mid-1980s, at about 1 million (references cited in the Preface, pp. 1-2, of The Origins, Persistence and Failings of HIV/AIDS Theory)—or maybe less, if you use numbers from the National Household Survey, above, rather than from the Centers for Disease Control and Prevention.
Perhaps, then, those 42,000 disappearing mysteriously each year represent people who eventually revert to HIV-negative after once having tested HIV-positive?
But of course the orthodox view is that seroreversion is exceedingly rare.
On the other hand, the evidence is that seroreversion is far from rare—see HIV “INFECTION” DISAPPEARS SPONTANEOUSLY, 22 January 2008).
With deaths from AIDS or “HIV disease”, you can also pick just about any number:
The Centers for Disease Control and Prevention likes to be on the safe side with its estimates. Where the National Vital Statistics Report (above) states 13,063 or 12,953 reported HIV/AIDS deaths for 2004, the Centers for Disease Control and Prevention estimates 18,099 for the same year (HIV/AIDS Surveillance Report, volume 17, revised June 2007).
Leave aside this discrepancy between 13,000 and 18,000 and consider only the methodology for a moment. It indicates that I should moderate my criticism, above, about reporting 13,063 and 12,953 when all that’s relatively reliably known is “about 13,000”. I had been tempted to repeat it concerning the estimate of 18,099 which should obviously have been given instead as “about 18,000”. But then it occurred to me that these estimates originate in computer programs, whose capabilities stretch to a much larger number of digits than a mere 5. Quite likely the computer spat out not 18,099 but something like 18,098.783. Instead of criticizing the resident experts, I should congratulate them for rounding up the estimate to the nearest person.