Re: Could the increase in human life-span account for these results?


"Paul Rubin" <rubin@xxxxxxx> wrote in message
news:3LUOl.15618$jZ1.12339@xxxxxxxxxxxxxxxxxxxxxxx
Gary wrote:
Below is an abstract of an article on apparent ageism in the choice of
photographs to accompany obituries - it seems (given the weakness of
measureing the apparent age of the person in the photograph) that
obituries are bing accompanied by ever younger portrait photographs.
But people have been living longer during the period of the study.
Could the greater life-span of people in the present day account for
the results (there being more years to choose from) without recourse
to ageism?


I don't think so -- the greater life spans add years to the right end of
the time line, not to the left end. If photos were being selected
randomly from a collection taken over a person's life span, extended spans
would result in a higher proportion of shots from late in life, not a
lower proportion.

/Paul

But (as I read it) the article isn't saying (explicitly) that the apparent
ages in the photographs got younger, it's saying that the _difference_
between photograph age and death age increased (actually, that the
proportion of cases in which that difference exceeded 15 years increased).

If the average length of the 'life line' increases, and (hypothetically)
photograph age is taken randomly from the life line (or from a particular
proportion of it), the size of that difference would also increase - the
whole thing just scales up.

<rant>
That said (and based only on the newspaper report and the university press
release http://researchnews.osu.edu/archive/obitphoto.htm, so I may be
wrong) I think we're idly speculating about causes of effects that may not
be real - as does the original work!

a) The data generation was pretty poor. This all depends on Anderson's
assessment of age.
<quote>Anderson conducted an exercise before doing the study in which he
estimated the ages of people in photographs for which he could confirm the
actual age of those shown. Within the 15-year window, he was accurate more
than four out of five times.</quote>
So the estimation was approx 20% "age-inaccurate", and we aren't told the
numbers there, so we've no estimate of uncertainty on that.
This makes his observed 'age-inaccuracy' changes from 17% to 27% to 30% to
36% seem less substantial; and what were the confidence intervals on those
estimates? and the differences?
Further, Anderson knew the purpose of both the main and preliminary
exercise, so was subject to (possibly unintentional) bias.
The accuracy of Andersons estimates may vary with attributes of the
photograph/person. If he underestimates older ages more (or less) than
younger, or women more (or less) than men, or photos from one decade more
(or less) than photos from another decade, these get totally confounded with
the effects he's deducing.

So I remain to be convinced there is a real effect to be explained.

b) Even if there is a real effect, his attributions of cause are totally
speculative.
"we were less accepting of aging in the 1990s than we were back in the 60s"
"Aging is a double whammy for women, who get hit with more ageism and
sexism,"
"how we as a society define these peak years, and how that definition has
changed over time"
"Adult children are thinking they want a picture of Dad when he was at his
best - and, especially in the late 1990s, that was significantly younger
than we he died. And the discrepancy was even larger for women,"
"Ageism seems to be increasing over time, despite our growing awareness of
the issue,"
He doesn't explain how these conclusions follow from the observations, or
why those putative causes are to be preferred to others, such as the one you
suggest (increasing life span), and the one I suggest (he can't reliably
estimate ages from photographs).

Overall, and in a statistics newsgroup, I regard this as an excellent
example of how NOT to plan, execute, and interpret an investigation.
</rant>

KJ


.

Relevant Pages