October 31 st , 2003

October 31st, 2003 October 31st, 2003

Finish Quasi-experiment True-experiment Lit review assignments

Full text onlySee reference lists of relevant articles (may require a trip to the library)See keywords of relevant articles

Realistic example Realistic example

Nonequivalent control Rehab vs. no rehab conditions

NCR O X O

NnoCR O O Q: Changes in reactivity to stressor May be unethical to randomize Assignment based on preexisting treatment

Realistic example Realistic example Nonequivalent control 2 classrooms of students

N O X ON O O

Behavioral intervention Randomize which classroom gets treatment Need separate classes to avoid resentment,

migration, etc.

Realistic example Realistic example

Other uses CHD and no CHD comparison group

NCHD O Xex O

NnoCHD O Xex O How does exercise influence?:

VO2 max change

Self-efficacy for exercise

True-experimental True-experimental

Take 2 identical situations Change an element in one situation What happens? E.g., Compare 2 cameras

True-experimental True-experimental

Goal here is to demonstrate:1. If X, then Y2. If not X then not Y

Does X ‘cause’ Y? Create identical groups of people Random assignment is key to success!

True-experimental True-experimental 3 characteristic of the true experiment help rule

out most threatsOne or more control groups AND one or more treatment groupsRandom assignment from population of interestVariable(s) of interest can be manipulated directly

True-experimental True-experimental

Two group post test only Better than previous designs Control group Assume randomization ensures baseline

equivalence

R X OR O

True-experimental True-experimental

Example X1 70% VO2max training X2 40% VO2max training O is fitness

R X1 OR X2 OR O

Multiple group post-test onlyMultiple group post-test only Advantage

Good for assessing differences after treatmentNo threat from a pre-testRelatively inexpensive

DisadvantagesSelection-MortalityE.g., lack of pain treatment,

or painful treatmentDon’t have a direct measure of change

True-experimental True-experimental

Classic pretest-posttest control group design Does treatment group change more than control

group? Group by time interaction

R O X OR O O

True-experimental True-experimental

Example: X1: 10,000 steps every day for 6 mos. X2: ‘Traditional’ activities for 6 mos. Control: remain on the couch O: VO2max, Blood lipids, glucose tolerance,

blood counts, etc.

R O X1 OR O X2 OR O O

The ClassicThe Classic Advantage

Good for assessing changes after treatmentMany threats ‘controlled’ (testing, maturation, etc.)Looks good on CV

DisadvantagesPretestSelection-MortalityE.g., lack of pain treatment,

or painful treatmentDiffusion of treatmentResentment for no treatment

Interpreting outcomesInterpreting outcomes Outcome 1: Both groups

increase Placebo effect? Maturation, etc. BUT…

Pretest Posttest

Control

Treat

Baseline

Interpreting outcomesInterpreting outcomes Outcome 2: Demonstrated

treatment effect Ideal case Interaction: Treatment by time

Pretest Posttest

Control

Treat

Interpreting outcomesInterpreting outcomes Outcome 3: Main effect over time Both groups improve Treatment ineffective

Pretest Posttest

Control

Treat

Interpreting outcomesInterpreting outcomes Outcome 4: No difference

between groups or over time

Pretest Posttest

Control

Treat

Solomon 4-group design Solomon 4-group design

Two Treatment, two no treatment One of each is pretested Used to control for testing threats Does test influence post-test score

R O X OR O OR X OR O

Trochim (1999)

Trochim (1999)

The ClassicThe Classic Disadvantages

Requires 4 groups of participantsCostlyDifficult to analyze

Next time: Advanced/ more complex designs.