Home Read Sign

in Search in book …

Want to create or adapt books like this? Learn more about how Pressbooks supports open
publishing practices.

SOCIAL DATA ANALYSIS

CONTENTS

Quantitative Data Analysis

9. Presenting the Results of

Quantitative Analysis
MIKAILA MARIEL LEMONIK ARTHUR

This chapter provides an overview of how to present the results of quantitative analysis, in
particular how to create effective tables for displaying quantitative results and how to write
quantitative research papers that effectively communicate the methods used and findings
of quantitative analysis.

Writing the Quantitative Paper

Standard quantitative social science papers follow a specific format. They begin with a title
page that includes a descriptive title, the author(s)’ name(s), and a 100 to 200 word abstract
that summarizes the paper. Next is an introduction that makes clear the paper’s research
question, details why this question is important, and previews what the paper will do. After
that comes a literature review, which ends with a summary of the research question(s)
and/or hypotheses. A methods section, which explains the source of data, sample, and vari-
ables and quantitative techniques used, follows. Many analysts will include a short discus-
sion of their descriptive statistics in the methods section. A findings section details the find-
ings of the analysis, supported by a variety of tables, and in some cases graphs, all of which
are explained in the text. Some quantitative papers, especially those using more complex
techniques, will include equations. Many papers follow the findings section with a discus-
sion section, which provides an interpretation of the results in light of both the prior litera-
ture and theory presented in the literature review and the research questions/hypotheses. A
conclusion ends the body of the paper. This conclusion should summarize the findings, an-
swering the research questions and stating whether any hypotheses were supported, par-
tially supported, or not supported. Limitations of the research are detailed. Papers typically
include suggestions for future research, and where relevant, some papers include policy im-
plications. After the body of the paper comes the works cited; some papers also have an
Appendix that includes additional tables and figures that did not fit into the body of the pa-
per or additional methodological details. While this basic format is similar for papers re-
gardless of the type of data they utilize, there are specific concerns relating to quantitative
research in terms of the methods and findings that will be discussed here.

Methods

In the methods section, researchers clearly describe the methods they used to obtain and
analyze the data for their research. When relying on data collected specifically for a given
paper, researchers will need to discuss the sample and data collection; in most cases,
though, quantitative research relies on pre-existing datasets. In these cases, researchers
need to provide information about the dataset, including the source of the data, the time it
was collected, the population, and the sample size. Regardless of the source of the data, re-
searchers need to be clear about which variables they are using in their research and any
transformations or manipulations of those variables. They also need to explain the specific
quantitative techniques that they are using in their analysis; if different techniques are used
to test different hypotheses, this should be made clear. In some cases, publications will re-
quire that papers be submitted along with any code that was used to produce the analysis
(in SPSS terms, the syntax files), which more advanced researchers will usually have on
hand. In many cases, basic descriptive statistics are presented in tabular form and ex-
plained within the methods section.

Findings
The findings sections of quantitative papers are organized around explaining the results as
shown in tables and figures. Not all results are depicted in tables and figures—some minor
or null findings will simply be referenced—but tables and figures should be produced for all
findings to be discussed at any length. If there are too many tables and figures, some can
be moved to an appendix after the body of the text and referred to in the text (e.g. “See
Table 12 in Appendix A”).

Discussions of the findings should not simply restate the contents of the table. Rather, they
should explain and interpret it for readers, and they should do so in light of the hypothesis
or hypotheses that are being tested. Conclusions—discussions of whether the hypothesis or
hypotheses are supported or not supported—should wait for the conclusion of the paper.

Creating Effective Tables

When creating tables to display the results of quantitative analysis, the most important
goals are to create tables that are clear and concise but that also meet standard conven-
tions in the field. This means, first of all, paring down the volume of information produced in
the statistical output to just include the information most necessary for interpreting the re-
sults, but doing so in keeping with standard table conventions. It also means making tables
that are well-formatted and designed, so that readers can understand what the tables are
saying without struggling to find information. For example, tables (as well as figures such as
graphs) need clear captions; they are typically numbered and referred to by number in the
text. Columns and rows should have clear headings. Depending on the content of the table,
formatting tools may need to be used to set off header rows/columns and/or total
rows/columns; cell-merging tools may be necessary; and shading may be important in ta-
bles with many rows or columns.

Here, you will find some instructions for creating tables of results from descriptive, crosstab-
ulation, correlation, and regression analysis that are clear, concise, and meet normal stan-
dards for data display in social science. In addition, after the instructions for creating tables,
you will find an example of how a paper incorporating each table might describe that table
in the text.

Descriptive Statistics

When presenting the results of descriptive statistics, we create one table with columns for
each type of descriptive statistic and rows for each variable. Note, of course, that depending
on level of measurement only certain descriptive statistics are appropriate for a given vari-
able, so there may be many cells in the table marked with an — to show that this statistic is
not calculated for this variable. So, consider the set of descriptive statistics below, for occu-
pational prestige, age, highest degree earned, and whether the respondent was born in this
country.
 R’s occupational prestige score (2010) Age of respondent
Valid 3873 3699
N
Missing 159 333
Mean 46.54 52.16
Median 47.00 53.00
Std. Deviation 13.811 17.233
Variance 190.745 296.988
Skewness .141 .018
Std. Error of Skewness .039 .040
Kurtosis -.809 -1.018
Std. Error of Kurtosis .079 .080
Range 64 71
Minimum 16 18
Maximum 80 89
25 35.00 37.00
Percentiles 50 47.00 53.00
75 59.00 66.00

Statistics

Frequency Percent Valid

Percent
Cumulative
Percent
R’s highest degree less than high
school 246 6.1 6.1 6.1
Valid 4009
N high school 1597 39.6 39.8 46.0
Missing 23
Median 2.00 associate/junior 370 9.2 9.2 55.2
Valid college
Mode 1
bachelor’s 1036 25.7 25.8 81.0
Range 4
graduate 760 18.8 19.0 100.0
Minimum 0
Total 4009 99.4 100.0
Maximum 4
Missing System 23 .6
Statistics Total 4032 100.0

R’s highest degree

 Was r born in this Frequency Percent Valid Percent Cumulative Percent
country yes 3516 87.2 88.8 88.8
Valid 3960 Valid no 444 11.0 11.2 100.0
N
Missing 72 Total 3960 98.2 100.0
Mean 1.11 Missing System 72 1.8
Mode 1 Total 4032 100.0
Statistics Was r born in this country

Table 1. SPSS Ouput: Selected Descriptive Statistics

To display these descriptive statistics in a paper, one might create a table like Table 2. Note
that for discrete variables, we use the value label in the table, not the value.

Occupational Age Highest Degree Born in This

Prestige Score Earned Country?
Mean 46.54 52.16 — 1.11
Median 47 53 1: Associates (9.2%) 1: Yes (88.8%)

Mode — — 2: High School —

(39.8%)
Standard 13.811 17.233 — —
Deviation
Variance 190.745 296.988 — —
Skewness 0.141 0.018 — —
Kurtosis -0.809 -1.018 — —
Less than High
Range 64 (16-80) 71 (18-89) School (0) – —
Graduate (4)
Interquartile 35-59 37-66 — —
Range
N 3873 3699 4009 3960

Table 2. Descriptive Statistics

If we were then to discuss our descriptive statistics in a quantitative paper, we might write
something like this (note that we do not need to repeat every single detail from the table, as
readers can peruse the table themselves):

This analysis relies on four variables from the 2021 General Social Survey: occupa-
 tional prestige score, age, highest degree earned, and whether the respondent was
born in the United States. Descriptive statistics for all four variables are shown in
Table 2. The median occupational prestige score is 47, with a range from 16 to 80.
50% of respondents had occupational prestige scores scores between 35 and 59. The
median age of respondents is 53, with a range from 18 to 89. 50% of respondents are
between ages 37 and 66. Both variables have little skew. Highest degree earned
ranges from less than high school to a graduate degree; the median respondent has
earned an associate’s degree, while the modal response (given by 39.8% of the re-
spondents) is a high school degree. 88.8% of respondents were born in the United
States.

Crosstabulation

When presenting the results of a crosstabulation, we simplify the table so that it highlights
the most important information—the column percentages—and include the significance
and association below the table. Consider the SPSS output below.
 R’s subjective class identification
lower working middle upper Total
class class class class
Count 65 106 68 7 246
less than high % within R’s
school subjective class 18.8% 7.1% 3.4% 4.2% 6.2%
identification
Count 217 800 551 23 1591
high school % within R’s
subjective class 62.9% 53.7% 27.6% 13.9% 39.8%
identification
Count 30 191 144 3 368
R’s highest associate/junior % within R’s
degree college subjective class 8.7% 12.8% 7.2% 1.8% 9.2%
identification
Count 27 269 686 49 1031
bachelor’s % within R’s
subjective class 7.8% 18.1% 34.4% 29.5% 25.8%
identification
Count 6 123 546 84 759
graduate % within R’s
subjective class 1.7% 8.3% 27.4% 50.6% 19.0%
identification
Count 345 1489 1995 166 3995
Total % within R’s
subjective class 100.0% 100.0% 100.0% 100.0% 100.0%
identification
Table 3. R’s highest degree * R’s subjective class identification Crosstabulation

Value df Asymptotic Significance (2-sided)

Pearson Chi-Square 819.579a 12 <.001
Likelihood Ratio 839.200 12 <.001
Linear-by-Linear Association 700.351 1 <.001
N of Valid Cases 3995
a. 0 cells (0.0%) have expected count less than 5. The minimum expected count is 10.22.
Chi-Square Tests
 Value Asymptotic Approximate Approximate
Standard Errora Tb Significance
Interval by Pearson’s R .419 .013 29.139 <.001c
Interval
Ordinal by Spearman .419 .013 29.158 <.001c
Ordinal Correlation
N of Valid Cases 3995
a. Not assuming the null hypothesis.
b. Using the asymptotic standard error assuming the null hypothesis.
c. Based on normal approximation.
Symmetric Measures

Table 4 shows how a table suitable for include in a paper might look if created from the
SPSS output in Table 3. Note that we use asterisks to indicate the significance level of the
results: * means p < 0.05; ** means p < 0.01; *** means p < 0.001; and no stars mean p > 0.05
(and thus that the result is not significant). Also note than N is the abbreviation for the
number of respondents.

Respondent’s Subjective Class Identification

Lower Class Working Middle Upper Class Total

Class Class
Less than 18.8% 7.1% 3.4% 4.2% 6.2%
High School
High School 62.9% 53.7% 27.6% 13.9% 39.8%
Highest Associate’s /
Degree Junior College 8.7% 12.8% 7.2% 1.8% 9.2%
Earned
Bachelor’s 7.8% 18.1% 34.4% 29.5% 25.8%
Graduate 1.7% 8.3% 27.4% 50.6% 19.0%
N: 3995 Spearman Correlation 0.419***

If we were going to discuss the results of this crosstabulation in a quantitative research pa-
per, the discussion might look like this:

A crosstabulation of respondent’s class identification and their highest degree

earned, with class identification as the independent variable, is significant, with a
Spearman correlation of 0.419, as shown in Table 4. Among lower class and working
class respondents, more than 50% had earned a high school degree. Less than 20%
of poor respondents and less than 40% of working-class respondents had earned
 more than a high school degree. In contrast, the majority of middle class and upper
class respondents had earned at least a bachelor’s degree. In fact, 50% of upper class
respondents had earned a graduate degree.

Correlation

When presenting a correlating matrix, one of the most important things to note is that we
only present half the table so as not to include duplicated results. Think of the line through
the table where empty cells exist to represent the correlation between a variable and itself,
and include only the triangle of data either above or below that line of cells. Consider the
output in Table 5.

R’s Highest year R’s family

Age of occupational of school R income in
respondent prestige score completed 1986 dollars
(2010)
Pearson 1 .087** .014 .017
Correlation
Age of respondent Sig. (2-tailed) <.001 .391 .314
N 3699 3571 3683 3336
Pearson .087** 1 .504** .316**
R’s occupational Correlation
prestige score Sig. (2-tailed) <.001 <.001 <.001
(2010)
N 3571 3873 3817 3399
Pearson .014 .504** 1 .360**
Correlation
Highest year of
school R completed Sig. (2-tailed) .391 <.001 <.001
N 3683 3817 3966 3497
Pearson .017 .316** .360** 1
Correlation
R’s family income in
1986 dollars Sig. (2-tailed) .314 <.001 <.001
N 3336 3399 3497 3509
**. Correlation is significant at the 0.01 level (2-tailed).
Table 5. SPSS Output: Correlations

Table 6 shows what the contents of Table 5 might look like when a table is constructed in a
fashion suitable for publication.
 Occupational Highest Year of Family Income in
Age Prestige Score School 1986 Dollars
Completed
Age 1
Occupational 0.087*** 1
Prestige Score
Highest Year of
School 0.014 0.504*** 1
Completed
Family Income in 0.017 0.316*** 0.360*** 1
1986 Dollars
Table 6. Correlation Matrix

If we were to discuss the results of this bivariate correlation analysis in a quantitative paper,
the discussion might look like this:

Bivariate correlations were run among variables measuring age, occupational pres-
tige, the highest year of school respondents completed, and family income in con-
stant 1986 dollars, as shown in Table 6. Correlations between age and highest year of
school completed and between age and family income are not significant. All other
correlations are positive and significant at the p<0.001 level. The correlation between
age and occupational prestige is weak; the correlations between income and occu-
pational prestige and between income and educational attainment are moderate,
and the correlation between education and occupational prestige is strong.

Regression

To present the results of a regression, we create one table that includes all of the key infor-
mation from the multiple tables of SPSS output. This includes the R2 and significance of the
regression, either the B or the beta values (different analysts have different preferences
here) for each variable, and the standard error and significance of each variable. Consider
the SPSS output in Table 7.

Model R R Square Adjusted R Square Std. Error of the Estimate

1 .395a .156 .155 36729.04841
a. Predictors: (Constant), Highest year of school R completed, Age of respondent, R’s occupational
prestige score (2010)
Table 7. SPSS Output: Regression
Model Sum of Squares df Mean Square F Sig.
Regression 805156927306.583 3 268385642435.528 198.948 <.001b
1 Residual 4351948187487.015 3226 1349022996.741
Total 5157105114793.598 3229
a. Dependent Variable: R’s family income in 1986 dollars
b. Predictors: (Constant), Highest year of school R completed, Age of respondent, R’s occupational
prestige score (2010)
ANOVAa

Unstandardized Standardized Collinearity

Coefficients Coefficients Statistics
Model t Sig.
B Std. Beta Tolerance VIF
Error
(Constant) -44403.902 4166.576 -10.657 <.001
Age of 9.547 38.733 .004 .246 .805 .993 1.007
respondent
R’s occupational
1 prestige score 522.887 54.327 .181 9.625 <.001 .744 1.345
(2010)
Highest year of
school R 3988.545 274.039 .272 14.555 <.001 .747 1.339
completed
a. Dependent Variable: R’s family income in 1986 dollars

Coefficientsa

The regression output in shown in Table 7 contains a lot of information. We do not include
all of this information when making tables suitable for publication. As can be seen in Table
8, we include the Beta (or the B), the standard error, and the significance asterisk for each
variable; the R2 and significance for the overall regression; the degrees of freedom (which
tells readers the sample size or N); and the constant; along with the key to p/significance
values.
 Beta & SE

Age 0.004
(38.733)
Occupational Prestige 0.181***
Score (54.327)
Highest Year of School 0.272***
Completed (274.039)
R2 0.156***
Degrees of Freedom 3229
Constant -44,403.902
* p<0.05 **p<0.01 ***p<0.001
Table 8. Regression Results for Dependent Variable
Family Income in 1986 Dollars

If we were to discuss the results of this regression in a quantitative paper, the results might
look like this:

Table 8 shows the results of a regression in which age, occupational prestige, and
highest year of school completed are the independent variables and family income
is the dependent variable. The regression results are significant, and all of the inde-
pendent variables taken together explain 15.6% of the variance in family income. Age
is not a significant predictor of income, while occupational prestige and educational
attainment are. Educational attainment has a larger effect on family income than
does occupational prestige. For every year of additional education attained, family
income goes up on average by $3,988.545; for every one-unit increase in occupa-
tional prestige score, family income goes up on average by $522.887.[1]

Exercises

1. Choose two discrete variables and three continuous variables from a dataset of
your choice. Produce appropriate descriptive statistics on all five of the vari-
ables and create a table of the results suitable for inclusion in a paper.
2. Using the two discrete variables you have chosen, produce an appropriate
crosstabulation, with significance and measure of association. Create a table of
the results suitable for inclusion in a paper.
3. Using the three continuous variables you have chosen, produce a correlation
matrix. Create a table of the results suitable for inclusion in a paper.
 4. Using the three continuous variables you have chosen, produce a multivariate
linear regression. Create a table of the results suitable for inclusion in a paper.
5. Write a methods section describing the dataset, analytical methods, and vari-
ables you utilized in questions 1, 2, 3, and 4 and explaining the results of your
descriptive analysis.
6. Write a findings section explaining the results of the analyses you performed in
questions 2, 3, and 4.

1. Note that the actual numberical increase comes from the B values, which are shown in
the SPSS output in Table 7 but not in the reformatted Table 8. ↵

LICENSE

Social Data Analysis Copyright © 2021 by Mikaila Mariel Lemonik Arthur and Roger Clark is
licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International
License, except where otherwise noted.

Powered by Pressbooks

Guides and Tutorials | Pressbooks Directory | Contact