1 Experimental Designs Pretest-Posttest Design Between Subjects

Experimental Designs
n Basic Designs
– Pretest-Posttest Design
– Between Subjects Designs
• Independent Groups Design
• Matched Groups Design
– Within Subject Designs (Repeated Measures)
Pretest-Posttest Design
n Used to measure effectiveness of a treatment (IV)
n Obtain measure of DV before treatment -- pretest
n Expose subjects to treatment
n Obtain measure of DV after treatment -- posttest
n If pretest and posttest measures are different
– Attribute change to treatment (IV)
n Ex: Measure effectiveness of a SAT preparation course
– Pretest measure of performance
– Take course
– Posttest measure of performance
Between Subjects Designs
n Assign different subjects to different conditions
– Receive different levels of IV
n Compare group means to determine if IV had an effect
n Types
– Independent Groups Design
– Matched Groups Design
Independent Groups Design
n Subjects are randomly assigned to various conditions/groups -- receive different levels of IV
n Compare mean performance for two groups
– If different -- attribute to IV
n Ex: Measure effect of caffeine on problem solving speed
– 50 subjects
• 25 randomly assigned to control group (decaf)
• 25 randomly assigned to experimental group
– Measure time needed to solve 10 spatial problems
– Compare mean of control group to mean of experimental group
Matched Groups Design
n Similar to Between-Groups Design
n Assign subjects to conditions by pairing or matching on important characteristic
– Important characteristic -- variable correlated (predictive of) DV
n Ex: Examine two different types of study techniques
– Outlining vs. question generation
– 50 subjects -- 25 assigned to each condition
– Match according to GPA
– Randomly assign one from each pair to the 2 conditions
Within-Subject Design
n AKA: Repeated Measures Design
n Each subject -- exposed to all conditions (all levels of IV)
n Comparing performance in different conditions
– Comparing performance of same people
n Ex: Memory for words
– compare concrete and abstract nouns (2 conditions)
– Subjects exposed to lists of concrete and abstract
– Test memory for both types of words
• Measure DV under both conditions
• Repeated measures
Pretest-Posttest Potential Problems
n Mortality
– Loss of subjects between pre and post
– Problem if loss changes sample characteristics
Pretest-Posttest Potential Problems
n History
– Intervening event between pretest and posttest
– Change due to event, not treatment
Pretest-Posttest Potential Problems
n Maturation // Spontaneous Remission
– Naturally occurring change between pre and post
– Would have occurred without treatment
Pretest-Posttest Potential Problems
n Regression to the mean
– Problem when treatment group selected based on extreme pretest score
– Some extreme scores due to “error”
– Next measurement -- closer to true score
Pretest-Posttest Potential Problems
n Testing Effects
– Exposure to the first test (pretest) alters response to second test
Ways to handle pretest-posttest problems
n Control Group -- pretest ------------- posttest
n Expected pattern for:
– Mortality
– History
– Maturation // Spontaneous Recovery
– Regression to the mean
– Testing Effects
Extraneous Variables and Confounds
n Extraneous variable
– Variable other than IV than has an effect on DV
– Not of interest in current experiment
– Ex: Parking lot territoriality (Study 2)
– IVs:
• status of intruding car
• Intrusion level (honk or not honk)
– DV:
• Departure time
– Extraneous variables -- other variables that might influence departure time
• Time of day, personality traits of driver, etc.
• Provide alternative explanation for results?
Extraneous Variables and Confounds
n Extraneous variables
– If equated across conditions -- not a problem to internal validity
– Simply add noise to data (more variability)
– More difficult to see effect of IV (if it exists)
n Confounds
– Extraneous variable that covaries with IV
– Provides alternative explanation for results
n Ex: Parking lot territoriality (Study 2)
– Variables that covaried with
• Status (car type)
• Intrusion (honking)
Between-Subjects Designs
n Main concerns
– Creation of comparable groups
• Avoid any systematic differences between groups
• Random assignment or matching should handle
– If done correctly
– Equivalent treatment of groups
• Groups treated same except for IV
• Care in setting up and conducting experiment
– More on experimental control next section
Within-Subject Designs
n Carryover / Sequence effects
– Subjects being exposed to all conditions
– Exposure to one condition - influence response to other conditions
– Practice Effects / Learning
– Fatigue Effects
– Contrast Effects
– Handle by using counterbalancing techniques
n Complete counterbalancing
n Incomplete counterbalancing
– Block randomization
– Latin square
Complete Counterbalancing
n Use all possible orderings of conditions
n Equal number of subjects -- each order
n # of sequences = N! (where N is # of conditions)
– N * N-2 * N-2 * N-3 etc.
n OK when # conditions is relatively small (4 or less)
n Use with larger # of conditions
n Present the conditions in a random order across subjects
n Order effects should balance-out with a sufficient number of subjects
Latin Square
n Use with larger number of conditions
n Each condition appears an equal number of times in each “temporal position”
n Each condition follows every other condition an equal number of times
Latin Square
Latin Square Sequence