27.10.2014 Views

Russel-Research-Method-in-Anthropology

Russel-Research-Method-in-Anthropology

Russel-Research-Method-in-Anthropology

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

Bivariate Analysis: Test<strong>in</strong>g Relations 617<br />

There are several ways that pairs of observations can be distributed if they<br />

are ranked on two ord<strong>in</strong>al variables.<br />

1. They can be ranked <strong>in</strong> the same order on both variables. We’ll call these ‘‘same.’’<br />

2. They can be ranked <strong>in</strong> the opposite order on both variables. We’ll call these<br />

‘‘opposite.’’<br />

3. They can be tied on either the <strong>in</strong>dependent or dependent variables, or on both.<br />

We’ll call these ‘‘ties.’’<br />

In fact, <strong>in</strong> almost all bivariate tables compar<strong>in</strong>g ord<strong>in</strong>al variables, there are<br />

go<strong>in</strong>g to be a lot of pairs with tied values on both variables. Gamma, written<br />

G, is a popular measure of association between two ord<strong>in</strong>al variables because<br />

it ignores all the tied pairs. The formula for gamma is:<br />

G <br />

No. of same-ranked pairs No. of opposite-ranked pairs<br />

No. of same-ranked pairs No. of opposite-ranked pairs<br />

Formula 20.13<br />

Gamma is an <strong>in</strong>tuitive statistic; it ranges from 1.0 (for a perfect negative<br />

association) to 1.0 (for a perfect positive association), through 0 <strong>in</strong> the middle<br />

for complete <strong>in</strong>dependence of two variables.<br />

If there are just two ord<strong>in</strong>al ranks <strong>in</strong> a measure, and if the number of opposite-ranked<br />

pairs is 0, then gamma would equal 1. Suppose we measure<br />

<strong>in</strong>come and education <strong>in</strong> a Malaysian community ord<strong>in</strong>ally, such that: (1)<br />

Anyone with less than a high school diploma is counted as hav<strong>in</strong>g low education,<br />

and anyone with at least a high school diploma is counted as hav<strong>in</strong>g high<br />

education; and (2) Anyone with an <strong>in</strong>come of less than RM34,000 a year is<br />

counted as hav<strong>in</strong>g low <strong>in</strong>come, while anyone with at least RM34,000 a year is<br />

counted as hav<strong>in</strong>g high <strong>in</strong>come. (RM is the symbol for the Malaysian r<strong>in</strong>ggit.<br />

RM34,000, or about $9,000 U.S. dollars, was the average per capita GDP <strong>in</strong><br />

Malaysia <strong>in</strong> 2005.)<br />

Now suppose that no one who had at least a high school diploma earned<br />

less than RM34,000 a year. There would be no pair of observations, then, <strong>in</strong><br />

which low <strong>in</strong>come and high education (an opposite pair) co-occurred.<br />

If the number of same-ranked pairs is zero, then gamma would equal 1.0.<br />

Suppose that no one who had high education also had a high <strong>in</strong>come. This<br />

would be a perfect negative association, and gamma would be 1.0. Both<br />

1.0 and 1.0 are perfect correlations.<br />

Calculat<strong>in</strong>g the Pairs for Gamma<br />

The number of same-ranked pairs <strong>in</strong> a bivariate table is calculated by multiply<strong>in</strong>g<br />

each cell by the sum of all cells below it and to its right. The number

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!