CONCEPTUALIZING AN AVERAGE 185
Measuring Individuals
Consider taking the height of 100 randomly chosen adult men in the United
States. Is the mean or median of these observations a central tendency? If so, what
does it represent? Many statisticians view the mean in this case as something like
the actual or true height of males in the United States (or in some subgroup). But
what could a statement like that mean?
For several reasons, an average in this situation is harder to view as a central
tendency than the average in the repeated measurement example. First, the gold
nugget and its mass are both perceivable. We can see and heft the nugget. In
contrast, the population of men and their average height are not things we can
perceive as directly. Second, it is clear why we might want to know the weight of
the nugget. But why would we want to know the average height of a population of
men? Third, the average height may not remain fixed over time, because of factors
such as demographic changes or changes in diet. Finally, and perhaps most
important, we cannot easily compartmentalize the height measurements into signal
and noise. It seems like a conceptual leap to regard each individual height as partly
true height, somehow determined from the average of the population, and partly
random error determined from some independent source other than measurement
error.
For all of these reasons, it is hard to think about the average height of the group
of men as a central tendency. We speculate, however, that it is somewhat easier to
regard differences between the averages of two groups of individual measurements
as central tendencies. Suppose, for example, we wanted to compare the average
height of U.S. men to the average height of (a) U.S. women or (b) men from
Ecuador. We might interpret the difference between averages as saying something in
the first case about the influence of genetics on height and in the second, about the
effects of nutrition on height. When making these comparisons, we can regard the
difference in averages as an indicator of the “actual effect” of gender or of nutrition,
things that are easier to imagine wanting to know about even if they are difficult to
observe directly.10
Some support for this speculation comes from Stigler (1999), who claims that
Quetelet created his infamous notion of the “average man” not as a tool to describe
single distributions, but as a method for comparing them: “With Quetelet, the
essential idea was that of comparison—the entire point was that there were different
average men for different groups, whether categorized by age or nationality, and it
was for the study of the nature and magnitude of those differences that he had
introduced the idea” (p. 61). Although we concede that the notion of a “true” or
“actual” value is still a bit strained in these comparison cases, we believe that one
needs some approximation to the idea of true value to make meaningful
comparisons between two groups whose individual elements vary. To see why, let
us look more closely at the comparison of men versus women.
Suppose we compute a mean or median height for a group of U.S. men and
another for a group of U.S. women. Note that the act of constructing the hypothesis
that gender partly determines height requires us to conceive of height as a process
CONCEPTUALIZING AN AVERAGE 185
Measuring Individuals
Consider taking the height of 100 randomly chosen adult men in the United
States. Is the mean or median of these observations a central tendency? If so, what
does it represent? Many statisticians view the mean in this case as something like
the actual or true height of males in the United States (or in some subgroup). But
what could a statement like that mean?
For several reasons, an average in this situation is harder to view as a central
tendency than the average in the repeated measurement example. First, the gold
nugget and its mass are both perceivable. We can see and heft the nugget. In
contrast, the population of men and their average height are not things we can
perceive as directly. Second, it is clear why we might want to know the weight of
the nugget. But why would we want to know the average height of a population of
men? Third, the average height may not remain fixed over time, because of factors
such as demographic changes or changes in diet. Finally, and perhaps most
important, we cannot easily compartmentalize the height measurements into signal
and noise. It seems like a conceptual leap to regard each individual height as partly
true height, somehow determined from the average of the population, and partly
random error determined from some independent source other than measurement
error.
For all of these reasons, it is hard to think about the average height of the group
of men as a central tendency. We speculate, however, that it is somewhat easier to
regard differences between the averages of two groups of individual measurements
as central tendencies. Suppose, for example, we wanted to compare the average
height of U.S. men to the average height of (a) U.S. women or (b) men from
Ecuador. We might interpret the difference between averages as saying something in
the first case about the influence of genetics on height and in the second, about the
effects of nutrition on height. When making these comparisons, we can regard the
difference in averages as an indicator of the “actual effect” of gender or of nutrition,
things that are easier to imagine wanting to know about even if they are difficult to
observe directly.10
Some support for this speculation comes from Stigler (1999), who claims that
Quetelet created his infamous notion of the “average man” not as a tool to describe
single distributions, but as a method for comparing them: “With Quetelet, the
essential idea was that of comparison—the entire point was that there were different
average men for different groups, whether categorized by age or nationality, and it
was for the study of the nature and magnitude of those differences that he had
introduced the idea” (p. 61). Although we concede that the notion of a “true” or
“actual” value is still a bit strained in these comparison cases, we believe that one
needs some approximation to the idea of true value to make meaningful
comparisons between two groups whose individual elements vary. To see why, let
us look more closely at the comparison of men versus women.
Suppose we compute a mean or median height for a group of U.S. men and
another for a group of U.S. women. Note that the act of constructing the hypothesis
that gender partly determines height requires us to conceive of height as a process
การแปล กรุณารอสักครู่..
