Transcript survey research

Chapter 2
Research Methods
by Neil Guppy
Copyright © 2004 by Nelson, a division of Thomson
Canada Limited.
1
SOCIAL RESEARCH
• Social research involves the systematic study
of the social world.
• Good social research aims to better the human
condition and enlarge the scope of human
freedom by requiring that social phenomena
be described, explained, understood, and
acted upon.
Copyright © 2004 by Nelson, a division of Thomson Canada Limited.
2
TYPES OF KNOWLEDGE
• Common sense is knowledge gained from
personal experience.
• Religious faith/mysticism is knowledge based on
spiritual teachings.
• Expertise is specialized knowledge from
experts.
Copyright © 2004 by Nelson, a division of Thomson Canada Limited.
3
SCIENTIFIC KNOWLEDGE I
• Scientific knowledge is based on direct,
systematic, reproducible observation. It is
the only type of knowledge that rests on
the principles of public scrutiny of
methods and results, scepticism, and
criticism.
Copyright © 2004 by Nelson, a division of Thomson Canada Limited.
4
SCIENTIFIC KNOWLEDGE II
• Scientific claims are falsifiable propositions.
• Although hunches, creativity, and imagination are
important in formulating scientific ideas, the rules
of scientific method require objectivity, or
minimizing personal bias.
• Social scientists differ from natural scientists by
studying people’s meaningful action; social
scientists’ subjects can think and act.
Copyright © 2004 by Nelson, a division of Thomson Canada Limited.
5
METHODS OF SOCIAL
RESEARCH I: EXPLANATION
• A fundamental premise underlying social research is
that correlation does not prove causation.
• Correlation can be spurious – that is, factors other
than the apparent cause of a phenomenon may be
responsible for it.
• An explanation is necessary to show how or why a
cause has a certain effect.
• Multiple causes are involved in almost every socialscientific explanation.
Copyright © 2004 by Nelson, a division of Thomson Canada Limited.
6
METHODS OF
SOCIAL RESEARCH II:
UNDERSTANDING
• Explanations must address the meaningfulness
of human activity. This is achieved through
understanding a relationship so as to clarify it.
• “Taking the role of the other” is a basic social
process by which other points of view can be
appreciated, understood, or reflected upon.
• Together, understanding and explanation
contribute to knowledge and social change.
Copyright © 2004 by Nelson, a division of Thomson Canada Limited.
7
VARIABLES I
Term
Definition
Example
Variable
An empirical object
or phenomenon that
can assume a
number of different
values.
Individual
annual
income
Copyright © 2004 by Nelson, a division of Thomson Canada Limited.
8
VARIABLES II
Term
Definition
Example
Dependent
A factor whose
value is assumed to
be caused or
brought about by
the operation of an
independent
variable.
Annual income
is a dependent
variable since
it is influenced
by number of
hours worked
per year.
variable
Copyright © 2004 by Nelson, a division of Thomson Canada Limited.
9
VARIABLES III
Term
Definition
Independent A factor that is
assumed to be
variable
responsible for
causing or bringing
about the value of
some other factor.
Copyright © 2004 by Nelson, a division of Thomson Canada Limited.
Example
The number of
hours a person
works in a year
is an
independent
variable that
influences the
person’s
annual income.
10
CORRELATION DOES NOT
PROVE CAUSATION
Copyright © 2004 by Nelson, a division of Thomson Canada Limited.
11
TECHNIQUES OF SOCIAL
RESEARCH
•
Sociologists have developed a variety of
techniques to gather evidence.
•
Three of the most important are:
• experiments;
• survey research; and
• observation studies.
Copyright © 2004 by Nelson, a division of Thomson Canada Limited.
12
EXPERIMENTS I
• Experiments allow researchers to isolate
presumed causes more precisely than any other
method.
• Using randomization, people in an experiment are
assigned to experimental and control groups
based on chance. This ensures few differences
between the two groups.
• The hypothesis is an unverified, testable claim
that is tested by assessing the relationship
between variables.
Copyright © 2004 by Nelson, a division of Thomson Canada Limited.
13
EXPERIMENTS II
Limitations of sociological experiments:
• Many major social processes are not
amenable to laboratory testing.
• Ethical and practical problems limit the use
of experiments.
• When people know they are being studied
they can become self-conscious and
change their behaviour accordingly (the
Hawthorne effect).
Copyright © 2004 by Nelson, a division of Thomson Canada Limited.
14
EXPERIMENTS III
• Field experiments are conducted in a natural
setting and therefore tend to reduce
problems of artificiality.
•
The degree to which the findings of
experimental research remain valid outside
the laboratory is known as their external
validity.
Copyright © 2004 by Nelson, a division of Thomson Canada Limited.
15
SURVEY RESEARCH I
•
The survey is the primary method of
collecting social science evidence.
•
It allows the systematic comparison of
answers to questions from a large sample, a
representative part of the population.
•
Researchers can then generalize results to
the larger population from which the sample
was drawn.
Copyright © 2004 by Nelson, a division of Thomson Canada Limited.
16
SURVEY RESEARCH II
•
Survey questions are posed through selfadministered questionnaires or through
personal interviews.
•
When conducting surveys, it is important not to
assume that people:
• understand what you are asking;
• know the answer to questions;
• will admit the answer to themselves; and
• will give valid answers.
Copyright © 2004 by Nelson, a division of Thomson Canada Limited.
17
OBSERVATION STUDIES I
• When using the observation method,
sociologists sometimes act as outside
observers and sometimes as participant
observers to see firsthand how people act.
• It is often difficult to gain access to actions
and events, particularly if they are private.
• Observation studies are especially useful for
understanding the meaning that actions have
for the people being investigated.
Copyright © 2004 by Nelson, a division of Thomson Canada Limited.
18
OBSERVATION STUDIES II
Problems with observation studies include:
• Presence: the chance that subjects may change
their behaviour due to the presence of the
observer.
• Ethnocentrism: the possible imposition of the
researcher’s values on the subject matter of the
study.
• Bias of focus: the possibility that responses and
findings are generated only as a result of the
questions being asked.
Copyright © 2004 by Nelson, a division of Thomson Canada Limited.
19
OTHER RESEARCH METHODS
• Sociologists often use historical evidence to
assess theories of social change.
• Sociologists also use official statistics,
collected by governments in censuses,
labour force surveys, police reports, etc., to
gain valuable information. These data,
however, are not “objective facts” and
should be interpreted with caution.
Copyright © 2004 by Nelson, a division of Thomson Canada Limited.
20
ANALYZING NUMERICAL DATA:
CONTINGENCY TABLES I
• Numerical data can be analyzed using contingency
tables (or cross-tabulations), which simultaneously
classify an individual’s scores on two or more
variables.
• To improve understanding of patterns in data, raw
data are converted into percentages.
• With percentages, one can see how the category an
individual occupies in the dependent variable is
affected by the category occupied in the independent
variable.
Copyright © 2004 by Nelson, a division of Thomson Canada Limited.
21
ANALYZING NUMERICAL DATA:
CONTINGENCY TABLES II
• The strength of association between two
variables in a contingency table can be measured
by the percentage difference between columns.
• By introducing a control variable in a two-variable
table, one can see how (if at all) the third variable
affects the percentage difference originally
observed. This clarifies the reasons for the
original association in the two-variable table.
Copyright © 2004 by Nelson, a division of Thomson Canada Limited.
22
THE RELATIONSHIP BETWEEN
LEVEL OF EDUCATION AND
EARNED INCOME
Insert Table 2.1, p. 48
Copyright © 2004 by Nelson, a division of Thomson Canada Limited.
23
THE RELATIONSHIP BETWEEN
LEVEL OF EDUCATION, EARNED
INCOME, AND GENDER
Copyright © 2004 by Nelson, a division of Thomson Canada Limited.
24
ANALYZING NUMERICAL DATA:
REGRESSION ANALYSIS I
• When analyzing data with multiple variables,
contingency tables often become complicated.
Regression analysis is a good alternative in
such cases.
• Data can be simultaneously cross-classified in
a scatterplot, with the independent variable
along the x-axis and the dependent variable
along the y-axis.
Copyright © 2004 by Nelson, a division of Thomson Canada Limited.
25
ANALYZING NUMERICAL DATA:
REGRESSION ANALYSIS II
A scatterplot can be
summarized statistically
using regression techniques. A straight line is
drawn so the distance from
each point to the line is
minimized. The further the
points are from the line, the
weaker the association
between the two variables.
Copyright © 2004 by Nelson, a division of Thomson Canada Limited.
26
SUPPLEMENTARY SLIDES
Copyright © 2004 by Nelson, a division of Thomson Canada Limited.
27
HOW RESEARCH
FILTERS PERCEPTION
Values
Theories
Previous Research
Methods
“Reality”
Copyright © 2004 by Nelson, a division of Thomson Canada Limited.
28
THE RESEARCH CYCLE
1. formulate question
6. report results
2. review existing
literature
5. analyze data
3. select method
4. collect data
Copyright © 2004 by Nelson, a division of Thomson Canada Limited.
29
CORRELATION
Dependent variable
Dependent variable
60
Dependent variable
60
60
r = .85
r = -.92
r=0
40
40
40
20
20
20
0
0
0
2
4
6
8
Independent Variable
1. Positive Correlation
0
0
2
4
6
8
Independent Variable
2. Negative Correlation
Copyright © 2004 by Nelson, a division of Thomson Canada Limited.
0
2
4
6
Independent Variable
3. No Correlation
30
8
THE MARGIN OF ERROR
IN A SAMPLE
2.5% margin of error
Support for
Candidate A
X
40%
48%
60%
2.5% margin of error
40%
Support for
Candidate B
X
50%
60%
In a sample of 1,500 people, 48% of the respondents support candidate
A and 50% support candidate B. However, because the 2.5% margins
of error overlap, we can’t be sure if support for candidates A and B
differs in the population. To conclude that support for the two candidates
differs in the population, the margins of error must not overlap.
Copyright © 2004 by Nelson, a division of Thomson Canada Limited.
31