Descriptive statistics for variables
Descriptive statistics for variables
QUANT/QUANT.
Table 1. Descriptive statistics for variables.
mean standard deviation Q1 Q2 Q3 min max n
educ 13.73 2.97 12 14 16 0 20 2345
paeduc 11.88 4.15 10 12 14 9 20 1687
… “The regression equation given by ordinary least squares (OLS) was educ = 0.29paeduc +
10.67. For a unit increase in the respondent’s father’s education – i.e., one year – we predict
that the respondent’s education increases by about ⅓ of a year, on average. Here, the intercept
is meaningful; for respondents whose fathers had zero formal education, their predicted level of
education was about 10 and ⅔ years. R-squared was 0.17, indicating that the ratio of variation
in the predictor (‘in the direction of’ the outcome) to variation in the outcome was 17 percent”.
Note: your situation may be more complex (e.g., your intercept may lack meaning). You
cannot simply swap out mine and Sanghyo’s numbers here!! You must think about your
own analysis!! Note that it is often a good idea, so long as you present the precise
numbers in the paper, to round them a small amount.
Figure 1.
QUAL/QUAL.
Table 1. Descriptive statistics for variables.
mean standard deviation n
protestant 0.49 0.5 2345
deathpen 0.63 0.48 2193
Table 2. Two-way table of protestant and deathpen.
Non-Protestant Protestant Total
Oppose 39.52 (441) 33.99 (361) 36.84 (802)
Favor 60.48 (675) 66.01 (701) 63.16 (1376)
Total 51.05 (1116) 48.95 (1139) 2178
… “The marginal distribution of religious affiliation – non-Protestant / Protestant – is about 51.05
to 48.95, while the marginal distribution support for the death penalty in the case of convictions
for murder is about 63.16 to 36.84 in favor. Protestants had a higher conditional probability of
supporting the death penalty: 66 percent to about 60.5 percent for non-Protestants.”
Figure 1.
QUANT pred / QUAL outcome
For the univariate analyses, just combine techniques from above as appropriate. For the bivariate
analysis, something like this is good… →
Table 1. Conditional means of income by religious affiliation.
Religious affiliation
Mean income (GSS
units)
Standard deviation of income (GSS
units) n
protestant 15.49 6.11 627
non-protestant 15.74 6.4 725
“Mean income – note that these units are not directly interpretable as true ratio numbers since
the GSS measures income only on an ordinal variable – was slightly higher for non-Protestants
than Protestants, though the difference was minor; the spread among Protestants was also
somewhat smaller”.