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.

618 Chapter 20<br />

of opposite-ranked pairs is calculated by multiply<strong>in</strong>g each cell by the sum of<br />

all cells below it and to its left. This is diagramed <strong>in</strong> figure 20.3.<br />

In Table 20.10,<br />

the number of same-ranked pairs is:<br />

16 (7 + 7 + 4 + 27) = 720<br />

+ 7 (7 + 27) = 238<br />

+ 7 (4 + 27) = 217<br />

+ 7 (27)<br />

= 189<br />

Total<br />

1,364<br />

Figure 20.3. Calculat<strong>in</strong>g gamma.<br />

The number of opposite-ranked pairs is:<br />

61 (7 + 7 + 13 + 4) = 1,891<br />

+ 7 (7 + 13) = 140<br />

+ 7 (4 + 13)<br />

+ 7 (13)<br />

=<br />

=<br />

119<br />

91<br />

Total<br />

2,241<br />

Gamma for table 20.10, then, is:<br />

G 1,3642,241<br />

1,3642,241 877<br />

3,605 .24<br />

Gamma tells us that the variables are associated negatively—people who<br />

agree with either of the statements tend to disagree with the other, and vice<br />

versa—but it also tells us that the association is relatively weak.<br />

Is Gamma Significant?<br />

How weak? If you have more than 50 elements <strong>in</strong> your sample, you can<br />

test for the probability that gamma is due to sampl<strong>in</strong>g error us<strong>in</strong>g a procedure<br />

developed by Goodman and Kruskal (1963). A useful presentation of the procedure<br />

is given by Loether and McTavish (1993:598, 609). First, the gamma<br />

value must be converted to a z-score, or standard score. The formula for convert<strong>in</strong>g<br />

gamma to a z-score is:<br />

z Gn s n o /2n1G 2 Formula 20.14<br />

where G is the sample gamma, is the gamma for the population, n is the size<br />

of your sample, n s is the number of same-ranked pairs, and n o is the number of<br />

opposite-ranked pairs.<br />

As usual, we proceed from the null hypothesis and assume that for the<br />

entire population is zero—that is, that there really is no association between<br />

the variables we are study<strong>in</strong>g. If we can reject that hypothesis, then we can<br />

assume that the gamma value for our sample probably approximates the<br />

gamma value, , for the population. Us<strong>in</strong>g the data from figure 20.3 and the<br />

gamma value for table 20.10:

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

Saved successfully!

Ooh no, something went wrong!