Biology Reference
In-Depth Information
PROBLEMS WITH STATURE ESTIMATION
Apples to Oranges: Data Comparison Problems between Antemortem
and Postmortem Data
When estimating stature in skeletal biology, you are essentially comparing a dry bone length
to the known stature of that individual. The way in which the living (or cadaver) stature was
measured or reported can vary greatly. This creates a scenario of trying to compare apples
(antemortem data) to oranges (postmortem data), which can be seriously fraught with error.
Snow and Williams (1971)
studied variations in living stature compared to skeletal remains.
They found stature discrepancies from four sources of data: (1) self-reported, (2) measured
with shoes, (3) not being fully erect, and (4) evening versus morning variation (
Bass, 1979
).
Ousley (1995)
elaborated on the two different types of errors when dealing with stature
estimation. Equations used can be flawed by incorrect measurement (postmortem) or
incorrect antemortem data collection. All osteometric and anthropometric studies providing
postmortem data are subject to inter- and intraobserver error. The antemortem data may also
have systematic bias due to interobserver error in measuring stature, misreporting on ID
cards, age changes in stature, and even due to such effects as the time of the day (you are taller
in the morning).
Ousley (1995)
describes some data as more precise and other data as more
accurate.
12
According to
Ousley (1995)
, the greater the precision, the narrower the range of
statures will be (i.e., there will be a low standard error), and the greater the accuracy, the
more likely the actual stature will be included within the range of error, which may require
widening that range. The goal for stature estimation ideally is both precision and accuracy.
Trotter and Gleser recommended broadening the error range to cover the 95% confidence
interval
,
because this will increase accuracy as the range is more likely to include the actual
stature, but the broader range will be less precise. Ousley recommends using what is called
a prediction interval to increase precision because, unlike standard error (SE), it accounts for
the sample size. This topic is discussed further in the chapter on age estimation by Uhl in this
volume (Chapter 3). Suffice it to say that for the best possible estimate of stature, Ousley still
suggests that the Fully method is the best estimator (
Ousley, 1995
).
Antemortem data can be recorded in two different ways that
Ousley (1995)
has defined as
(1) measured stature (MSTAT) and (2) forensic stature (FSTAT). Medical or military records
that include measurements of living stature (MSTAT) are extremely prone to interobserver
error.
Ousley (1995)
noted that variation in MSTAT can be as much as 5 inches (12.7 cm) in
some cases, depending on whether shoes were worn or even due to daily fluctuation in
stature.
13
FSTAT is easier to obtain for missing persons, as it is the stature that is self-reported
on the driver's license/identity card or as reported by a family member. FSTAT is subject to
systematic bias due to age-related height reduction and misreporting discrepancies between
males and females and between taller individuals and shorter individuals (
Giles and
Hutchinson, 1991; Willey and Falsetti, 1991
).
Giles and Hutchinson (1991)
used data from
a sample of 8000 military personnel and found that, depending on height, men overestimate
12
See Uhl (Chapter 3), this volume for further explanation of accuracy versus precision.
13
Stature taken early in the morning or immediately following a nap can be as much as two 2 cm taller
(
Ousley, 1995
).
