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 637<br />

his or her social network; but (3) The l<strong>in</strong>ear regression equation is hardly any<br />

better than the global mean at reduc<strong>in</strong>g error <strong>in</strong> predict<strong>in</strong>g the dependent variable.<br />

You can see this by compar<strong>in</strong>g the mean l<strong>in</strong>e and the regression l<strong>in</strong>e (the<br />

slightly diagonal l<strong>in</strong>e <strong>in</strong> figure 20.8). They are very similar.<br />

What that regression l<strong>in</strong>e depicts, of course, is the correlation between age<br />

and size of network, which is a puny –.099. But if we <strong>in</strong>spect the data visually,<br />

we f<strong>in</strong>d that there are a couple of natural ‘‘breaks.’’ It looks like there’s a break<br />

<strong>in</strong> the late 20s, and another somewhere <strong>in</strong> the 60s. We’ll break these data <strong>in</strong>to<br />

three age chunks from 12 to 26, 27 to 61, and 64 to 89, take separate means<br />

for each chunk, and see what happens. I have marked the three chunks and<br />

their separate means on table 20.19.<br />

Like r, eta must be squared to f<strong>in</strong>d the variance accounted for. Eta 2 is calculated<br />

from the follow<strong>in</strong>g formula:<br />

2 1<br />

yy c 2<br />

2<br />

Formula 20.24<br />

yy<br />

where y c is the average for each chunk and y is the overall average for your<br />

dependent variable. For table 20.19, eta-squared is:<br />

2 1 3,871.55<br />

10,058.55 .62<br />

which is the proportionate reduction of error <strong>in</strong> predict<strong>in</strong>g the number of<br />

friends people have from the three separate averages of their age, rather than<br />

from the global average of their age. This shows a pretty strong relation<br />

between the two variables, despite the very weak Pearson’s r.<br />

Statistical Significance, the Shotgun Approach, and Other Issues<br />

To f<strong>in</strong>ish this chapter, I want to deal with four thorny issues <strong>in</strong> social science<br />

data analysis: (1) measurement and statistical assumptions, (2) elim<strong>in</strong>at<strong>in</strong>g the<br />

outliers, (3) significance tests, and (4) the shotgun method of analysis.<br />

Measurement and Statistical Assumptions<br />

By now you are comfortable with the idea of nom<strong>in</strong>al-, ord<strong>in</strong>al-, and <strong>in</strong>terval-level<br />

measurement. This sem<strong>in</strong>al notion was <strong>in</strong>troduced <strong>in</strong>to social science<br />

<strong>in</strong> a classic article by S. S. Stevens <strong>in</strong> 1946. Stevens said that statistics like t

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

Saved successfully!

Ooh no, something went wrong!