Russel-Research-Method-in-Anthropology

572 Chapter 19
mortality rates above 110 per 1,000—Chad, Liberia, and Gambia—the mean
life expectancy for women rises 1.8% to 68.23 years, and for men it rises 1.7%
to 63.63 years. Chad, Liberia, and Gambia comprise just 0.002 (one-fifth of
1%) of the world’s population, but the data from those countries pull down
the mean life expectancy for women and men by a factor of about 9 (that is,
0.002% 9 1.8%). Median life expectancy, however, rises just 0.0025
(four-tenths of 1%) for women when we remove Chad, Liberia, and Gambia,
and it rises just 0.007 (seven-tenths of 1%) for men.
When interval data (like life expectancy) are normally distributed, the mean
is the best indicator of central tendency. When interval data are highly skewed,
the median is often a better indicator of central tendency. You absolutely must
get a feel for the shape of distributions to understand what’s going on.
Shape: Visualizing Distributions
A really good first cut at understanding whether data are normal or skewed
is to lay them out graphically. This is easy to do with any of the full-featured
statistics programs out there these days. I’ll show you six ways to lay out your
data: bar graphs and pie charts for nominal and ordinal variables; stem-andleaf
plots, box-and-whisker plots, histograms, and frequency polygons for
interval variables.
Bar Graphs and Pie Charts
Bar graphs and pie charts are two popular ways to graph the distribution
of nominal and ordinal variables. Figure 19.1 shows bar graphs for two of the
variables in table 19.2: GENDER and GUNGHO. Figure 19.2 shows the pie
charts for the same variables. Notice that in the bar charts, the bars don’t touch
one another. This indicates that the data are nominal or ordinal and not continuous.
The categories of the variables are shown along the horizontal axis of the
bar graph. The horizontal, or x-axis, is also called the abscissa. The number
of each category is shown on the left vertical axis. The vertical, or y-axis, is
also called the ordinate. You can, of course, show the percent of each category
on the y-axis instead.
In figure 19.1a, men are labeled 1 and women are labeled 2. Notice that it
makes no difference whether we put the bar for men or the bar for the women
on the left or the right when we graph GENDER. There is no order implied in
the attributes of a nominal variable. When we graph ordinal variables, however,
like GUNGHO, the order of the bars becomes important.