RESEARCH DESIGN

1
CHAPTER 4

• RESEARCH DESIGN

2
Research design

• A set of decisions regarding
• (1) what topic is to be studied,
• (2) among what population,
• (3) with what research methods,
• (4) for what purpose.

3
I. What topic is to be studied?

• What is your research question? What social
  phenomenon are you investigating? (It must be
  possible to do research on the research
  question.)
• What units will you study?
• Units of analysis - what or whom will be studied
  those things we examine in order to create
  summary descriptions of all such units and to
  examine differences and similarities among them

4
Types of units of analysis

• 1. Individuals
• 2. Groups
• 3. Organizations
• 4. Social Artifacts
• a. Social objects
• b. Social interactions
• 5. Geographic units (not in text)

5
1. Individuals e.g., females, union members,
Democrats, lesbians, faculty members, single
parents

• General Social Survey (most variables)
• Abortion attitudes
• Vote for President in 1996
• Highest educational degree

6
(No Transcript)
7
2. Groups - e.g., families, juvenile gangs,
friendship groups, married couples.

• General Social Survey
• Total family income last year
• Household composition
• U.S. Department of Justice, Office of Juvenile
  Justice and Delinquency Prevention - 1996 Youth
  Gang Survey

8
(No Transcript)
9
(No Transcript)
10
3. Organizations - formal social organizations,
e.g., universities, corporations, armies, churches

• Bureau of Justice Statistics - U.S. Jail
  statistics
• Census Bureau - Annual Survey of Manufactures

11
4. Social Artifacts - any product of social
beings or their behavior

• a. Social objects - e.g., books, songs, laws,
  jokes, graffiti, student excuses for missing
  exams, editorials
• Graffiti (Sherri Cavan)
• b. Social interactions - e.g., weddings,
  funerals, fights, court cases
• Heavyweight title fights in New York

12
5. Geographic units - e.g., countries, census
tracts, states, city blocks, neighborhoods

• Countries as units of analysis World Factbook
• Variables on which countries can be compared
• U.S. States as units of analysis U.S. Census
  Bureau
• Choose a state, then a county.
• Virginia counties in 2000 census

13
II. Among what population?

• Define exactly which population we are
  studying.FBI Uniform Crime Reports (FAQs What
  is NIBRS?)
• U.S. Department of Labor - 1995 Survey of
  Employer Provided Training (see Technical Note
  about halfway down the page)
• Bureau of Labor Statistics - Employment
  Characteristics of Families (Technical Note)
• Most social scientific research is not conducted
  with entire populations, but rather with samples.

14
III. With what research methods?

• 1. Conceptualization and operationalization (Ch.
  5).
• 2. Data collection methods experiments (Ch. 8),
  survey research methods (Ch. 9), secondary data
  analysis (Ch. 9), field research methods (Ch.
  10), unobtrusive methods (Ch. 11).
• 3. Decisions about the time dimension

15
3. Decisions about the time dimension

• a. Cross-sectional design a study based on
  observations representing a single point in time
• Example 2000 General Social Survey
• b. Longitudinal design a study design involving
  the collection of data at different points in time

16
Types of longitudinal designs (1)

• 1. Trend study a longitudinal study in which a
  given characteristic of some population is
  monitored over time
• Example 1972-2000 General Social Surveys Analyze
  (menu bar) Action - Frequencies or
  crosstabulation Row - abany, Column - year,
  Percentaging - Column)
• Try others e.g., abhlth, coneduc
• Example U.S. Census Bureau - U.S.A. Statistics
  in Brief

17
Types of longitudinal designs (2)

• 2. Cohort study a longitudinal study in which
  subdivision of a population (a "subpopulation")
  is studied over time. Often, but not always,
  conducted on subpopulations that are defined by a
  time grouping. A cohort study can also be a panel
  study (see below).
• Example Social Statistics Research Unit, City
  University (Great Britain) (both cohort and panel)

18

• Example of a cohort study that is NOT a panel
  study General Social Survey data
• Go to our class directory on the L drive
  (L\Faculty\jhonnold\SOCY POLI 320-SPRING04)
• Double click on - cohort analysis syntax.sps
• In SPSS, click Run All
• Go to output window to view analysis

19
Types of longitudinal designs (3)

• 3. Panel study a longitudinal study in which
  data are collected from the same sample (the
  panel) at several points in time
• Example Panel Study of Income Dynamics (PSID)
• Example HRS and AHEAD (See 2)

20
4. Decisions about data analysis techniques

• 1. Qualitative data analysis (Ch. 13)
• 2. Quantitative data analysis (Ch. 14)

21
IV. For what purpose?

• 1. Research motivations
• (1) Pure
• (2) Applied
• 2. Research purposes
• (1) Exploration
• (2) Description
• (3) Explanation

22
1. Research motivations

• (1) Pure research - research conducted solely for
  the purpose of advancing social scientific
  knowledge "basic" research.
• Example National Science Foundation
• (2) Applied research - research undertaken for
  the purpose of influencing some phenomenon in the
  "real" world research conducted for the purpose
  of putting the results into practice
• Examples of Applied Research from the Institute
  for Research on Poverty
• Examples from the Institute for Women's Policy
  Research (click on "Our Research")
• The Social Science Research Council funds efforts
  blending basic and applied research.

23
2. Research purposes

• (1) Exploration research conducted for the
  purpose of providing a beginning familiarity with
  a topic, to test the feasibility of undertaking a
  more extensive study, or to develop methods to be
  employed in a more extensive study
• (2) Description research conducted for the
  purpose of providing descriptive information
  about a social phenomenon or group
• (3) Explanation research undertaken for the
  purpose of explaining relationships among social
  scientific variables devoted to understanding
  causes of social phenomena and answering the
  "why" question

24

• Most studies will have elements of two or three
  research purposes.
• All research design elements should be formulated
  with the purpose(s) and the research topic(s) or
  question(s) of the study firmly in mind.

25
The logic of causation in explanatory social
scientific research

• Two different models of social scientific
  explanation (understanding/discovering causes)
• 1. Idiographic
• 2. Nomothetic

26
1. Idiographic Model

• Explains a social phenomenon through the
  enumeration of the numerous, perhaps unique,
  idiosyncratic causal factors that lie behind that
  particular condition or event. Researcher
  attempts to understand all of the causes of a
  particular social phenomenon.

27
Illustration of Idiographic Model

• Why did you come to VCU instead of going to ODU?
• Why else?
• What are some more reasons?
• Name so many reasons until you come to the point
  that anyone with all those reasons would have to
  have come to VCU.

28
2. Nomothetic Model

• Explains a social phenomenon through the
  discovery of relatively few causal factors that
  lie behind a class of conditions or events.
  Researcher aims to find the fewest factors with
  the greatest explanatory power a probabilistic
  approach.

29
Illustration of Nomothetic Model-1

• What are the main reasons VCU students chose VCU
  over ODU?
• What would be two or three variables that would
  enable us to make a prediction about whether
  students will choose VCU over ODU?
• However, we couldnt predict perfectly.

30
Illustration of Nomothetic Model-2

• Research question - why do some people have
  better access to health care than others?
• Finding - people with low incomes are less likely
  to receive timely outpatient care for medical
  conditions and more likely to have
  hospitalizations that could have been avoided by
  timely care.

31
Would we say the above nomothetic explanation is
false if

• We find that low income isn't the only factor
  related to timely receipt of outpatient care?
• We can't explain all instances of untimely
  outpatient care in terms of low income?
• Some people in low-income areas do receive timely
  outpatient care or some people in high-income
  areas do not?
• Only a minority of people in low-income areas do
  not receive timely outpatient care?

32
False Criteria for Nomothetic Causality

• 1. Complete Causation - In order to be considered
  cause(s) of a phenomenon, a factor or set of
  factors must completely explain the phenomenon.
• INCORRECT - one or more factors may be incomplete
  explanations for the phenomenon and be considered
  causes.

33

• 2. Exceptional Cases - In order to be considered
  a valid causal explanation, no exceptions to the
  explanation are allowed.
• INCORRECT - exceptions can exist the explanation
  is probabilistic, not deterministic.
• 3. Majority of Cases - An explanation must apply
  to the majority of the cases to be considered a
  valid causal explanation.
• INCORRECT. Cases sharing a variable attribute
  (people with low incomes) must be more likely to
  show a result (not have timely ambulatory care)
  than cases classified as sharing another
  attribute (people with high incomes).

34
Criteria for Nomothetic Causality

• 1. Time order - The cause must precede the effect
  in time (time order criterion).
• 2. Correlation - The two variables must be
  empirically correlated with one another
  (correlation criterion). But note that
  "correlation is not causation.
• 3. Nonspurious - The observed empirical
  correlation cannot be explained in terms of some
  third variable that causes both of the other two
  (nonspuriousness criterion). More generally,
  there can be no alternative explanation to the
  effect of the independent variable.

35
Hypothetical Example of Spurious Relationship (p.
87)
36
Nomothetic Causality Criteria- Application

• In a study of women in colleges and universities,
  500 women were selected at random for each of
  three editions of Whos Who of American Women
  (total of 1500 in the sample). The 1100 college
  graduates in the sample were analyzed further.
  Results In comparison women who graduated
  from coeducational colleges and universities,
  women who graduated from womens colleges were
  about twice as likely to achieve high levels of
  occupational success. The womens colleges not
  only graduated about twice the number of female
  high achievers per 1000 students, but also
  employed about twice the number of female faculty
  per 1000 students. ConclusionThe authors
  concluded that womens colleges were superior to
  coeducational institutions in nourishing the
  talents of young female students.

37
Is their conclusion justified?

• Task 1 - Identify the independent and dependent
  variables.
• Task 2 - Can you conclude that the independent
  variable is a cause of the dependent variable in
  this situation? In other words, are the three
  criteria for nomothetic causality (below) met? If
  yes, why? If no, why not?
• Time order - the cause must precede the effect in
  time.
• Correlation - the two variables must be
  empirically correlated with one another.
• Nonspuriousness - the observed empirical
  correlation cannot be explained in terms of some
  third variable that causes both of the other two.

38
Perfectly Necessary and Sufficient Causes

• Necessary Cause - Must be present for the effect
  to occur
• Sufficient Cause - Always results in the effect
• Necessary and Sufficient Cause - Must be present
  for the effect to occur and always results in the
  effect

39
Necessary Cause (Must be present for the effect
to occur)
40
Sufficient Cause (Always results in the effect)
41
Necessary and Sufficient Cause (Must be present
for the effect to occur and always results in the
effect)
42
Recall

• Nomothetic explanation is probabilistic.
• Causes need NOT be perfectly necessary and
  sufficient, perfectly necessary, or perfectly
  sufficient in order to contribute to the
  explanation of a social phenomenon.
• However, the stronger the correlation between
  independent and dependent variables, assuming
  satisfaction of the other two nomothetic criteria
  (time order, nonspurious), the better the causal
  explanation.