ANALYSIS OF QUANTITATIVE DATANOTES
- Practical advice on coding and handling quantitative
data comes from survey research. See discussions in
Babbie (1998:366–372), Backstrom and Hursh-Cesar
(1981:309–400), Fowler (1984:127–133), Sonquist and
Dunkelberg (1977:210–215), and Warwick and Lininger
(1975:234–291). - Note that coding gender as 1 Male, 2 Female, or
as 0 Male, 1 Female, or reversing the gender for
numbers is arbitrary. The only reason one uses numbers
instead of letters (e.g., M and F) is that many computer
programs work best with all numbers. Sometimes cod-
ing data as a zero can create confusion, so the number 1
is usually the lowest value. - For discussions of many different ways to display
quantitative data, see Fox (1992), Henry (1995), Tufte
(1983, 1991), and Zeisel (1985:14–33). - Other statistics measure special types of means for
ordinal data and for other special situations, which are
beyond the level of discussion in this book. - On the elaboration paradigm and its history, see Bab-
bie (1998:400–409) and Rosenberg (1968).
6. Beginning students and people outside the social sci-
ences are sometimes surprised at the low (10 to 50 per-
cent) predictive accuracy in multiple regression results.
There are three responses to this. First, a 10 to 50 per-
cent reduction in errors is really not bad compared to
purely random guessing. Second, positivist social sci-
ence is still developing. Although the levels of accuracy
may not be as high as those of the physical sciences, they
are much higher than for any explanation of the social
world possible 10 or 20 years ago. Finally, the theoreti-
cally important issue in most multiple regression mod-
els is less the accuracy of overall prediction than the
effects of specific variables. Most hypotheses involve the
effects of specific independent variables on dependent
variables.
7. In formal hypothesis testing, we test the null hypoth-
esis and usually want to reject the null because rejection
of the null indirectly supports the alternative hypothesis
to the null, the one we deduce from theory as a tentative
explanation. The null hypothesis was discussed in
Chapter 6.