Title: Interpreting Correlations
1Program Evaluation Designs
2Programs for Class Discussion
- Counseling program for prison inmates
- Stress reduction training for employees
- Study skills for at-risk college students
- Parental training for problem children
3Why Should I Care About Research?
- Answering questions and making decisions
- We encounter research every day
- Common sense is often wrong
4While Youre on Vacation, Burglars Go to
Work According to FBI statistics, over 26 of
home burglaries take place between Memorial Day
and Labor Day. To help you protect your home, we
are offering a special summer sale price for
installation of a home security system. Home
Safety Experts Route 11 and Wayland Road 555-6676
5How Do I Know What to Research?
- Ideas
- Hypotheses - well thought-out suggestions or
ideas - Theories - systematic sets of assumptions
regarding the nature and cause of particular
events
6Idea Hypothesis Theory
It seems that people dont work as hard in a group as when they are alone When pulling on a rope, a person working by himself will exert more force than a person working in a group Sucker effect Free-rider Individual effort will not be noticed
Social Loafing Example
7Idea Hypothesis Theory
It seems that poor people are more violent than rich people There will be a correlation between income and the number of times arrested for being violent 1. 2. 3. 4. 5.
Aggression Example
8How Do I Find Previous Research?
- Written sources
- Journals
- Trade Magazines
- Magazines
- Books
- Electronic Resources
- First Search
- Psych Info
- InfoTrac
- The Web
9Where Will I Conduct Research?
- Locations
- Laboratory
- Field
- Office
- Issues
- External validity (generalizability)
- Control
10What Research Method Should I Use?Independent
and Dependent Variables
- Independent Variable
- Experimental group
- Control group
- Dependent Variable
11A researcher thinks that smaller groups will be
more cohesive than larger groups Independent
variable Group size Dependent variable Level
of cohesion
Number of Group Members
3 5 7 9 11 13
Cohesiveness rating 87 77 65 60 60 58
12A researcher thinks that asking for a large
request and then a smaller request will increase
the chance of a person agreeing to the smaller
request Independent variable Size of initial
request Dependent variable agreeing to small
request
Experimental Condition
Small request only Large request first
agreeing 18 79
13Identify the Independent and Dependent Variables
- Are people from large families more likely to
commit crimes than people from small families?
(A) - Will taking an SAT prep class increase scores on
the SAT? (B) - Will taking Vitamin A decrease night blindness?
(c)
14Identify the Independent and Dependent Variables
- A researcher found that a higher percentage of
serial killers have head injuries than do
nonviolent criminals (d) - A researcher found that people who set goals
perform better at work than people who do not set
goals (e) - A researcher found that the amount of time that
passes between viewing a crime and reporting a
crime influences the number of errors made by
eyewitnesses (f)
15Evaluation of Program ResultsCriteria
- Content validity
- Participant reactions
- Participant learning
- Behavioral change
- Bottom-line measures
16Dependent Measures in Program Evaluation
- Did people know about the program?
- Did people participate?
- Did they stay in the program?
17Dependent Measures in Program Evaluation
- Did they change?
- Reality versus perception
- Measures
- Attitudes
- Behavior
- Skills
- Health
- How much of a change?
- Any change
- Enough to meet some criterion
18Dependent Measures in Program Evaluation
- What did they think about the program
- Process
- Effectiveness
19What Research Method Should I Use?Research
Methods
- Experiment
- independent variable is manipulated and
- Subjects are randomly assigned to conditions
- dependent variable
- Quasi-experiment
- Independent variable is not manipulated or
- Subjects are not randomly assigned to conditions
- Case study
- Survey
- Interview
- Natural observation
20Evaluation of Program ResultsSingle-Group,
Nonexperimental Designs
Program
Posttest
21Evaluation of Program ResultsSingle-Group,
Nonexperimental Designs
Pretest
Program
Posttest
22Threats to Internal Validity
- Changes Unrelated to the Program
- Maturation
- History
- Participant Issues
- Selection
- Attrition
- Regression toward the mean
- Data Collection Issues
- Testing
- Instrumentation
23Construct Validity in Pretest-Posttest Designs
- Of concern when self-report measures are used
- Types of self-reported change
- Alpha (real change)
- Beta (respondents change their understanding of
the scale) - Gamma (respondents change their understanding of
the variable being measured) - Examples
- Asking students about school safety
- Asking couples about their marital satisfaction
24ExampleNew Child Care Center
- 2000 Employee absenteeism rate 5.09
- 2001 On-site child-care center established
(Jan 1) - 2001 Employee absenteeism rate 3.13
25Date Absenteeism External Factor Internal Factor
1/00 2.8
2/00 3.1
3/00 4.7 Unemployment rate at 4.1
4/00 4.7
5/00 4.8
6/00 6.7 Main highway closed
7/00 6.5
8/00 4.9 Highway reopens
9/00 4.5
10/00 4.4
11/00 8.7 Terrible snowstorm
12/00 5.3
2000 Total 5.09
26Date Absenteeism External Factor Internal Factor
1/01 5.3 Child care center started
2/01 5.2
3/01 5.1 Flextime program started
4/01 2.0 Unemployment rate at 9.3
5/01 2.0
6/01 2.0
7/01 1.8 Wellness program started
8/01 1.8
9/01 2.0 New attendance policy
10/01 2.1
11/01 4.0 Mild weather
12/01 4.2 Mild weather
2001 Total 3.13
27Demonstrating Effects in Designs Without a
Control Group
- Means
- Show means for the periods before and after
intervention - Use a t-test to determine statistical
significance - Report d to show practical significance
- (Mafter Mbefore) SDoverall
- Time Series Analysis
- Looks at trends over time
- Interrupted time series involves an intervention
- Most powerful is ABAB design
- Most common is AB design
28Child Care Center Example
Child Care Center Started
29Child Care Center Example
Child Care Center Started
30Child Care Center Example
Child Care Center Started
31Conformity SuicidePhillips (1980)
32Conformity and Commercial Airline
Fatalities(Phillips, 1983)
33Time Series Charts
- Include actual plotted data
- Include Lowess lines
- Smoothed data
- Plotted point is the median of the 3 adjacent
points - Should include notations for other important
events
34Control Group Designs
- Control for many of the threats to internal
validity - Groups need to be as similar as possible
- Using multiple groups increases accuracy of
interpretation - Placebos are often used for the control group
- Control group treatment
- Never get treatment
- Get treatment later
35Evaluation of Program ResultsControl Group
Designs
Pretest
Program
Posttest
Pretest
Posttest
36Evaluation of Program ResultsSolomon four-groups
design
Program
Posttest
Group 1
Pretest
Program
Posttest
Group 2
Pretest
Posttest
Group 3
Posttest
Group 4
37Assigning Subjects to Conditions
- Two Choices
- Independent groups
- Repeated measures
- Considerations
- Number of available subjects
- Order effects
- Will participating in one condition affect
responses in another
38Where Do I Get My Subjects?Who Will Participate?
- Size
- Students vs. real world
Does it Matter? If you were investigating whether
the type of evidence (e.g., eyewitness testimony,
DNA, fingerprints) influenced jurors decisions,
would students as subjects be different from
having people from the community?
39Where Do I Get My Subjects?Sampling
- Types of Samples
- Random
- Representative
- Non-random/representative
- Sampling Methods
- Random selection
- Simple
- Block randomization
- Proportionate
- Convenience
- Random assignment
40Randomized Block Designs
- Used when
- there are several conditions
- you want equal numbers in each condition
- You are concerned about time-related variables
- Example
- 4 conditions (A, B, C, D), 8 subjects per block
- Number of blocks the number of subjects desired
for each condition - Random blocks might be CBAD, DABC, BCDA, ACDB,
CEBA, DBCA, BCAD, ADBC,
41What Type of Sampling Method is Being Used?
- A researcher has the students in her classes fill
out a questionnaire - A researcher gives 6 to people who will
participate in his study. As the people arrive,
he flips a coin to see if they will be in the
experimental or the control condition. - A manager wants to see if a training program will
increase performance. She selects every third
name from the company roster to participate.
Employees with an odd number at the end of their
social security number are given one training
program and those with an even number are given
another.
42Where Do I Get My Subjects?Inducements to
Participate
- Extra credit
- Money
- Intrinsic reasons
- Ordered to participate
Does it Matter? Would the inducement used affect
the type of person agreeing to participate? In
what ways?
43Where Do I Get My Subjects?Informed Consent
- Ethically required
- Can be waived when
- Research involves minimal risk
- Waiver will not adversely affect rights of
participants - Research could not be done without the waiver
44Is Informed Consent Needed?
- An experimenter wants to study the effects of
electric shock on reducing patients depression
levels - A researcher wants to conduct a telephone survey
in which she asks people their five favorite TV
shows. She will then determine if males and
females like different shows. - A researcher wants to determine the types of
people who litter. He plans to hide above a road
and record information about the people who
litter or dont litter (e.g., age, sex, type of
car).
45Running the Study
- Informed consent
- Instructions
- Task completion
- Deception?
- Debriefing
46External Validity
- Can our findings be generalized to other
situations? - Considerations
- Sample
- How realistic was the task?
- How realistic was the dependent variable?
- Has the study been replicated in other situations?
47Putting it all Together
Redesign Your Study