E N D
1. Research Methods for the Social Sciences: An Introductory Course March 1st, 2010 – More on Design: Causality
Division on Addictions, Cambridge Health Alliance
Harvard Medical School
2. Agenda Revisiting Research Design
Experimental vs. Quasi-Experimental
Longitudinal Designs
Prevalence vs. Incidence / Individual Trajectories
Moderation and Mediation
Sample Studies
3. Caveat: Association Does Not Equal CausationCorrelate Does Not Equal Determinant
6. Correlation vs. Experiment Correlational research
Goal: To describe relationship between two or more variables.
NO CAUSALITY
Experimental research
Goal: Cause-and-effect relationship between variables JuliaJulia
7. Does Caffeine Increase Blood Pressure? WRONG Approach
Ask participants (or observe) how many cups of coffee they drink
Measure blood pressure
Calculate the correlation coefficient RIGHT Approach
Manipulate amount of coffee
Experimental group – drink
Control group – do not drink
Random group assignment
Test difference between groups JuliaJulia
8. Properties of an Experiment Manipulate at least 1 independent variable
Assign participants to the various conditions in a way that assures their initial equivalence
Control extraneous variables
JuliaJulia
9. Assignment of Participants to Conditions Simple random assignment (Between Group)
Every participant has equal probability of being placed in any experimental condition.
Matched Random assignment (Between Group)
Participants are assigned to groups based on their score on a relevant measure.
Repeated measures (Within group)
Same participant is assigned to different conditions. JuliaJulia
10. Control Extraneous Variables Eliminate or hold constant factors (confounds), other than IV that may affect the outcomes of the experiment.
If these factors are random, they create noise that can mask any effects
If they vary systematically with condition, they can create confounds JuliaJulia
11. Types of variable Dependent variables – OBSERVED
Independent variables – MANIPULATED
Control variables – held CONSTANT JuliaJulia
12. Quasi-Experiment
No IV manipulation.
No random group assignment.
13. Type 1. Pretest-Posttest Design O1 – pretest measure of students’ drug use
X – intervention
O2 – posttest measure of drug use.
14. Type 1. Pretest-Posttest Design.Weaknesses. Extraneous variables (History effect)
Something else affected the IV
Maturation effect
Students grew up.
Regression to the mean
The tendency for extreme scores to move, or regress toward the mean (because of measurement error).
Attrition
Some subjects may drop out before posttest.
15. Type 2. Simple Interrupted Time-Series Design O1 O2 O3 O4 X O5 O6 O7 O8
Taking several pretest measures of the DV before introducing an IV. And then, taking several posttest measures.
16. Type 2. Simple Interrupted Time-Series Design
17. Type 3. Longitudinal Designs O1 O2 O3 O4
IV: Time
18. Longitudinal Designs. Weaknesses. Attrition
Moving, lost motivation, mortality, etc.
Attrition can often introduce systematic bias
Researchers’ effort, time, money.
19. Evaluating Quasi-Experimental Designs. To conclude causality
The presumed causal variable preceded the effect in time.
The cause and the effect covary.
All other alternative explanations are eliminated.
20. Aggregate vs. Individual Trajectory Most of the information we collect tells us about the average behavior of groups, but not about individual patterns
Individual trajectory research explores the course of a behavior or disorder within an individual
21. Cross-Sectional design. Comparison between different age groups.
Weaknesses:
Generation effect
No individual development
22. When we look at prevalence rates, we often assume things about individuals…When we look at prevalence rates, we often assume things about individuals…
23. But it’s possible that with a stable prevalence rate, individuals still recoverBut it’s possible that with a stable prevalence rate, individuals still recover
24. Aggregate vs. Individual
25. Aggregate vs. Individual
26. Moderation and Mediation Once we have found an effect: X ? Y, we can ask two questions
What mediates the effect
What moderates the effect
27. What mediates the effect? Questions of mediation are questions of mechanism
How does X ? Y? Through what process?
28. What mediates the effect? Questions of mediation are questions of mechanism
How does X ? Y? Through what process?
29. What moderates the effect? Questions of moderation are questions of subgroup
For whom does X ? Y?
30. What moderates the effect? Questions of moderation are questions of subgroup
For whom does X ? Y?
31. How Can We Really Know Causality?
32. Oprah and Jenny McCarthy
33. From Oprah Interview “In recent years, the number of children diagnosed with autism has risen from 1 in every 500 children to 1 in 150—and science has not discovered a reason why. Jenny says she believes that childhood vaccinations may play a part. "What number will it take for people just to start listening to what the mothers of children who have seen autism have been saying for years, which is, 'We vaccinated our baby and something happened.”
34. From Oprah Interview “Jenny says even before Evan received his vaccines, she tried to talk to her pediatrician about it. "Right before his MMR shot, I said to the doctor, 'I have a very bad feeling about this shot. This is the autism shot, isn't it?' And he said, 'No, that is ridiculous. It is a mother's desperate attempt to blame something,' and he swore at me, and then the nurse gave [Evan] the shot," she says. "And I remember going, 'Oh, God, I hope he's right.' And soon thereafter—boom—the soul's gone from his eyes."
35. Freed et al., 2009 Online survey of parents of children up to age 17
1 in 4 parents agree that “some vaccines cause autism in healthy children”
36. Perceived Link Autism diagnoses have increased as MMR vaccines have increased
Autism develops around the same age as the MMR vaccine is given
Autism used to be confined to upper classes and did not cross class lines until vaccination did so
Symptoms of autism mirror symptoms of mercury poisoning
37. History of the MMR – Autism Link (reviewed by Gerber and Offit, 2009) Wakefield et al., 1998
12 Consecutively referred patients to pediatric GI department with pervasive developmental disorders
In 8 of these patients, parents or doctors identified the onset of problems as occurring immediately after MMR vaccination.
Gerber & Offit Analysis
Because of the timing of emergence of autism, and the number of children who develop autism, Gerber & Offit point out that the number of children Wakefield saw with autism developing shortly after MMR vaccination could occur by chance.
38. History of the MMR – Autism Link (reviewed by Gerber and Offit, 2009) Epidemiological Studies that followed
Rates of autism diagnosis did not increase significantly after introduction of MMR vaccine in 1987
No clustering of autism diagnoses around time of vaccination
Vaccinated and unvaccinated children just as likely to develop autism
There has been an increase in autism diagnosis from 1988 to 1999 though vaccination rates have remained stable.
39. Article 2: Hobfoll et al., 2009. Trajectories of resilience, resistance, and distress during ongoing terrorism: The case of Jews and Arabs in Israel Research Question:
What is the prevalence of different mental health trajectories in a population exposed to ongoing terrorism and what predicts those trajectories?
This is a question that attempts to identify psychosocial factors that influence mental health in a given context.
40. Article 2: Hobfoll et al., 2009. Trajectories of resilience, resistance, and distress during ongoing terrorism: The case of Jews and Arabs in Israel Trajectories
Resistant (no symptoms at T1 or T2)
Resilient (symptoms at T1 but not T2)
Delayed distress (symptoms at T1 but not T2)
Chronic distress (symptoms at both T1 and T2)
41. Article 2: Hobfoll et al., 2009. Trajectories of resilience, resistance, and distress during ongoing terrorism: The case of Jews and Arabs in Israel Hypotheses:
Greater economic resources will distinguish resistant and resilient individuals from distressed individuals.
Greater social support, less resource loss, less exposure, and less positive psychological change will distinguish resistant and resilient individuals from distressed individuals
Non-religious individuals will be more likely to belong to the resistant and resilient trajectories (compared to the distressed trajectory) than religious individuals.
Fewer resources and greater loss of resources will distinguish delayed distress individuals from resistant individuals.
42. Article 2: Hobfoll et al., 2009. Trajectories of resilience, resistance, and distress during ongoing terrorism: The case of Jews and Arabs in Israel Sample and Design:
Nationally representative telephone survey of adult Israelis, stratified by region (N = 1,613)
Initial response rate of 57% at T1 and 44% of the sample provided data at T2, so analyses are based on one quarter of potential respondents
43. Measures: Predictors
Demographics: Age, gender, income, education, ethnicity, religiousness, marital status
Terrorism exposure (# of types: 0,1,2+)
Loss of resources (economic or psychosocial)
Post-terrorism growth (e.g., hope, intimacy, etc.)
Social support
Measures: Outcome
PTSD symptoms
Depression symptoms
2+ PTSD or depressive symptoms used as the cut to classify participants as resistant or not at a given timepoint Article 2: Hobfoll et al., 2009. Trajectories of resilience, resistance, and distress during ongoing terrorism: The case of Jews and Arabs in Israel
44. Analysis Plan
Descriptives for each trajectory
Logistic Regressions
Resistant vs. Distressed
Gender, ethnicity, religiousness, income, education, psychosocial resource loss at T1 & T2, T2 social support
Resilient vs. Distressed
Ethnicity, income, T2 psychosocial resource loss, T2 post-traumatic growth
Delayed Distress vs. Resistant
Education, T2 psychosocial resource loss Article 2: Hobfoll et al., 2009. Trajectories of resilience, resistance, and distress during ongoing terrorism: The case of Jews and Arabs in Israel