Biology Reference
In-Depth Information
STATISTICAL APPROACHES
Metric traits are measured on a continuous scale (e.g., maximum cranial breadth can be,
at least theoretically, any value between 0 and
)whereasmorphoscopic traits are
measured categorically. The statistical treatment of continuous data has several advantages
over categorical data, which is assigned a value that is in one of several possible categories.
Categorical variables do not always have a numerical meaning. In other words, variables
like skin color (light, dark) and nasal aperture width (narrow, intermediate, wide) do not
necessarily have an explicit numerical equivalent. Because of this feature, categorical
data are not appropriately treated with the same statistical methods as their continuous
(i.e., metric, numerical) counterparts. The statistical methods used to evaluate continuous
data (see examples below) are more widely used and are generally better understood
than those used for categorical methods. Therefore, we will briefly outline the methods
used to treat continuous variables and spend the majority of our discussion on the treat-
ment of categorical data. As stated, this is a brief overview and several statistical concepts
will be introduced in the upcoming section. A certain knowledge of statistics is assumed
d
if
any of these concepts are unfamiliar, we recommend taking graduate level statistics courses
to catch up.
N
Metric Methods
Analyzing craniometric (and more recently postcranial metric) data to assess ancestry has
a long history in anthropology. To analyze craniometric data, the first and therefore most
important step is the proper collection of craniometric data. In the past, such data were
collected using sliding and spreading calipers, craniofor, and mandibulometers, among other
tools. However, a large number of laboratories are switching to three-dimensional digitizers
for data collection (e.g., see McKeown and Schmidt [Chapter 12], this volume). No matter the
method of data acquisition, the theoretical underpinnings are the same: the collection of land-
mark data and interlandmark distances for use in data analysis. The landmarks used by
forensic anthropologists are rooted in the earlier work of several prominent (though often
infamous) physical anthropologists
d
recall Morton's early craniometric data collection.
However,
Martin (1914)
and
Howells (1973, 1989, 1995)
are considered the “gold standards”
for landmark descriptions, illustrations, and definitions and should be consulted regularly by
both inexperienced and experienced anthropologists. Of course, reading the literature and
landmark definitions is no substitute for mentoring. Find an experienced anthropologist,
pester them to no end and watch over their shoulder as they explain the nuances of data
collection
d
it worked for us, it will work for you, as well. The interlandmark distances
used in a final analysis (such as FORDISC
d
discussed below) are outlined in
Martin
(1914)
;
Howells (1973, 1989, 1995)
and
Buikstra and Ubelaker (1994)
.
FORDISC and Discriminant Function Analysis
Once the data have been appropriately collected the next step is finding and using an
appropriate known reference sample. In the United States, this is most often the Forensic
Anthropology Databank (
Jantz and Moore-Jansen, 1988
) and the computer program
