Research Design - School of Journalism and Mass Communication

Download Report

Transcript Research Design - School of Journalism and Mass Communication

Research Design
Main Tasks of Research Design
• Specifying what you want to find out: this
involves explaining the concepts you are
interested in and how they will be measured
(explication and meaning analysis).
• Determining the best way to do it: this involves
determining whom or what you will explore,
describe, or explain (unit of analysis); what time
dimension is appropriate for your observations;
and how you are going to do it (method).
Design Depends on Goals
 What is the purpose of the research?
 What shall we observe, among whom, for what
purpose, and in what time frame?
 Exploration
 Often where inquiry begins
 Informal methods, feasibility testing
 Description
 Detail the features of elements under review
 E.g., U.S. Census
 Explanation
 Usually the aim of social scientists
 Relating variables to account for process
Unit of Observation & Analysis
 Unit of observation
 What we look at to make observation
 E.g., People in a survey, articles in content analysis
 Unit of analysis
 What we are interested in studying
 Usually same as unit of analysis; sometimes different:
E.g., Are “traditional” marriages more successful?
Unit of observation: husbands, wives
Unit of analysis: marriage type (couple)
Types of Units of Analysis
 Individuals
 Groups
 Organizations
 Social Interactions
 Social Artifacts
- However, bear in mind this is only one typology.
- E.g., Lofland’s: practices, episodes, encounters, roles, relationships, groups,
organizations, settlements, social worlds, lifestyles, and subcultures.
- What is important is the logic of units of analysis.
Types of Units of Analysis
 Individuals
 Groups
 Organizations
 Social Interactions
 Social Artifacts
- However, bear in mind this is only one typology. E.g., Lofland’s:
practices, episodes, encounters, roles, relationships, groups,
organizations, settlements, social worlds, lifestyles, and subcultures.
- What is important is the logic of units of analysis.
Units of Analysis: Individuals
 Most common unit of analysis in social
science/mass communication research
 Seek to explain differences between individuals
and relationships among individual differences
 Variables and Relationships:
 E.g., Income, Age, Gender, Education
 E.g., Associated with differences in tolerance
Units of Analysis: Social Groups
 Examples:
 Households, families, neighborhoods, gangs
 Seek to explain differences between groups and
relationships among those differences
 Variables:
 Households: income, media use (Nielsen)
 Marriages: types, communication patterns
 Neighborhoods: crime rates, income stratification
Units of Analysis: Organizations
 Examples:
 Corporations, Universities, Governments
Groups with formal organizational structures
Seek to explain differences between formal social
organizations and the relationships among
organizational differences
Variables:
 E.g., Corporations: employees, benefits, productivity
Units of Analysis: Social Interactions
 Examples:
 Kisses, Arguments, Email exchanges, discussion styles
 Social interaction are usually the product of
interplay between individuals.
 Studies seek to explain different types of social
interactions (ex. discussion as unit of analysis),
the types of people engaging in certain
interactions (ex. Individual as the unit of analysis)
 Variables:
 Number of arguments, argumentative people
Units of Analysis: Social Artifacts
 Examples:
 TV programs, newspaper articles, documents
 Social artifacts are any product of social beings or
their behaviors.
 Studies seek to explain differences between social
artifacts, the artifacts produced by different
source, and the relationships among these factors
 Variables:
 Level of violence, number of sources used
Faulty Reasoning & Units of Analysis
 Problems of drawing conclusions across units of
analysis:
 Ecological fallacy
 Reductionism
Ecological Fallacy
 Observed characteristic of group leads to:
 Inference about individual members
 Similar to Prejudice
 Individual judgments based on beliefs about group
 E.g., Precinct voting records are unit of observation
concerning support for democratic candidates
 Majority Black precincts vote democratic
 You can not assume that Blacks uniformly vote democratic
 Whites within precincts may be responsible for pattern
Reductionism
 Reducing complex phenomenon in a way that
privileges particular units of analysis over others
 E.g., Crime is a function of individual characteristics
 What about social structures?
 Economists: Economic reductionism
 Psychologists: Psychological reductionism
 Sociologists: Sociological reductionism
 Dominant paradigms often limit views
Time Dimension & Research Design
 Time and issues of causation
 Static designs:
 Cross-sectional study
 Longitudinal designs:
 Trend studies
 Cohort studies
 Panel studies
Cross-sectional Studies
 Static snapshot
 Slice of population at one point in time
 E.g., An opinion poll
 Inherent limitations:
 Inability to capture change over time
 Making causal inferences is dangerous
Cross-sectional studies
Top Global Concerns for 2003
% who list item among top 3 personal concerns
Source: Roper Reports Worldwide 2003
Study of 30,000 consumers age 13 to 65 in 30 countries
Cross-sectional studies
Longitudinal Designs
 Multiple observations across time
 Tracking changes across time
 Maybe in response to stimulus that occurs between
observations
 Testing for changes resulting from some intervening factor or
event
 Pretest-Posttest design in experimentation
Longitudinal Designs: Trends
 Measures change in population over time
 Sequential cross-sections of the population
 E.g., Changes over time in:
 Public knowledge levels
 Voter turnout rates
 Presidential approval ratings
 Inherent limitations:
 Starting point
 Inability to capture individual change over time
Longitudinal Designs: Trends
Longitudinal Designs: Cohort Studies
 Tracking changes in a group as they age
 E.g., People born in 1940 sampled every 10 years
 Measure change across the aging process
 E.g., Do people become more conservative?
 Cannot answer this question with a cross-sectional
design because differences in age may be due to cohort
or lifecycle differences.
Longitudinal Designs: Cohort Studies
Longitudinal designs: panel studies
 Goes a step further:
 Interviewing the same people more than once
 Captures change in individuals over time
 E.g., NES (cross-sectional and panel)
 How do people react over time?
 E.g, Public health/info campaigns
 The respondent mortality problem
 Are those who drop out different?
Longitudinal designs: panel studies
Statement:
“Most people are
honest”
On a six point
scale ranging
from definitely
disagree to
definitely agree.
Source: Life Style
Study – conducted
by Market Facts on
behalf of DDBChicago and
Dhavan V. Shah
Wave 1
Feb. 1999
N= 3,348
Wave 2
June 2000
N= 1,886
Wave 3
Nov. 2000
N= 1,282
Wave 4
July 2001
N= 964
Longitudinal designs: comparisons
Cross-sectional study
Trend study
2000
1990
2000
21-30
31-40
41-50
51-60
61-70
21-30
31-40
41-50
51-60
61-70
21-30
31-40
41-50
51-60
61-70
Cohort study
1990
21-30
31-40
41-50
51-60
61-70
2000
31-40
41-50
51-60
61-70
71-80
Panel study
2010
41-50
51-60
61-70
71-80
81-90
1990
21-30
31-40
41-50
51-60
61-70
2000
31-40
41-50
51-60
61-70
71-80
2010
41-50*
51-60*
61-70*
71-80*
81-90*