27.10.2014 Views

Russel-Research-Method-in-Anthropology

Russel-Research-Method-in-Anthropology

Russel-Research-Method-in-Anthropology

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

Multivariate Analysis 681<br />

4. Across all three replications, the three factors <strong>in</strong> tables 21.23 and 21.24 account<br />

for 70%–80% of the variance <strong>in</strong> the orig<strong>in</strong>al data matrix. That is, about threefourths<br />

to four-fifths of the variance <strong>in</strong> the orig<strong>in</strong>al data is accounted for by just<br />

three underly<strong>in</strong>g variables (the three factors) rather than the full list of orig<strong>in</strong>al<br />

variables. For women across the Caribbean, the construct of domestic cooperation<br />

is multidimensional and comprised of three subconstructs: shar<strong>in</strong>g of everyday<br />

domestic chores, shar<strong>in</strong>g of responsibilities for children, and affection from<br />

men as def<strong>in</strong>ed by be<strong>in</strong>g treated as an equal.<br />

Multidimensional Scal<strong>in</strong>g Analysis (MDS)<br />

MDS is another multivariate data-reduction technique. Like factor analysis,<br />

it is used to tease out underly<strong>in</strong>g relations among a set of observations. Also<br />

like factor analysis, MDS requires a matrix of measures of associations—e.g.,<br />

a correlation matrix based on th<strong>in</strong>gs like r, tau, gamma, etc. But unlike factor<br />

analysis, MDS does not require metric data.<br />

When you measure someth<strong>in</strong>g like how strongly people feel about someth<strong>in</strong>g,<br />

the numbers you assign to their feel<strong>in</strong>gs don’t have the same mean<strong>in</strong>g<br />

as, say, numbers that express distances <strong>in</strong> kilometers or kilograms of game<br />

meat killed per month. These latter numbers are called ‘‘metric’’ because they<br />

are grounded <strong>in</strong> well-understood units of measurement.<br />

Most attitude and cognition data are nonmetric. MDS is particularly useful<br />

for anthropologists, s<strong>in</strong>ce a lot of the measurements we make are nonmetric.<br />

Also, MDS produces a graphic display of the relation among any set of items,<br />

whether those items are people, or th<strong>in</strong>gs, or questions about attitudes. (For<br />

more about the theory beh<strong>in</strong>d MDS, see Romney et al. [1972].)<br />

How MDS Works<br />

Suppose you measure the association among three variables, A, B, and C,<br />

us<strong>in</strong>g Pearson’s r. The association matrix for these three variables is <strong>in</strong> the<br />

<strong>in</strong>side box of table 21.25.<br />

Clearly, variables A and C are more closely related to one another than are<br />

A and B, or B and C. You can represent this with a triangle, as <strong>in</strong> figure 21.3a.<br />

In other words, we can place po<strong>in</strong>ts A, B, and C on a plane <strong>in</strong> some position<br />

relative to each other. The distance between A and B is longer than that<br />

between A and C (reflect<strong>in</strong>g the difference between .50 and .80); and the distance<br />

between B and C is longer than that between A and C (reflect<strong>in</strong>g the<br />

difference between .40 and .80). The numbers <strong>in</strong> this graph are similarities:<br />

The lower the correlation, the longer the distance; the higher the correlation,<br />

the shorter the distance. (We won’t consider negative correlations here.)

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

Saved successfully!

Ooh no, something went wrong!