1/31
Looks like no tags are added yet.
Quasi-Experimental Research
Resembles experimental research, but is not true experimental research
-IV manipulation is present, but control group is missing or random assignment isn’t used
Conducted when random assignment is difficult or impossible
Types of Quasi-Experimental Designs
One-Group Designs
Non-Equivalent Group Designs
One Group Designs
Studies a single treatment group experiencing an experimental condition
-Outcome/DV is measured after intervention, or at multiple time points (before and after)
-Comparison happens within the SAME GROUP OVERTIME
No random assignment since there’s one group and no control/comparison
-Makes it more weak when providing causal explanation
Types of One-Group Designs
-One-group Posttest Only Design
-One-group Pretest-Posttest Only Design
-Interrupted Time Series Design
One-Group Posttest Only Design
Treatment is implemented (or IV manipulated), dependent variable is then measured once AFTER the treatment
-Very Weak: Insufficient evidence for causality, no way of knowing if outcome would have happened anyway
One-Group PreTest-Posttest Design
Single group has DV measured before and after treatment implemented
-Effect is inferred from difference between two scores
Weak to Moderate: Provides a “within-group comparison” but vulnerable to many internal validity threats (history, maturation, etc)
-Improvement over one-group posttest design
Internal Validity Threats in One-Group Designs
-History - Other things happened between pre/post test causing a change in DV scores
-Maturation - Participants changed between pre/posttest due to growth/learning
-Testing - Act of measuring the DV during pretest affects participant response on posttest
-Instrumentation - change of characteristics of instrument used to measure
-Regression to the Mean - Individuals who score extremely high/low tend to score less extremely on the next occasion
-Spontaneous Remission - Tendency for psychological problems to improve over time, with NO form of treatment
Interrupted Time Series Design
Set of measurements for one group taken at regular intervals overtime, which is interrupted by treatment/intervention
Moderate - Can account for pre-existing trends
-Remains vulnerable to coinciding events
Nonequivalent Group Design
Compare two or more groups: one experiences the introduced condition, and another doesn’t (or get a different one).
-Because people aren’t randomly assigned, the groups are likely to differ or be “non-equivalent” from the start
-Groups often come from already-established groups or from people’s own choices
Key Feature: There is a separate control or comparison group, but the group cannot be assumed to be equivalent before the intervention and may differ in ways other than the IV
Types of Nonequivalent Group Designs
Posttest Only Non-equivalent Groups
Pretest-Posttest Non-equivalent Groups
Interrupted Time-Series Non-Equivalent Groups (Control Series)
Pretest-Posttest with Switching Replication
Switching Replication w/ Treatment Removal
Posttest Only Nonequivalent Group Design
Two (or more) nonrandomly formed groups compared after one receives a condition. No pretest
Weal: Unsure if groups already different on DV before conditions (due to lack of pretest)
Pretest-Posttest Nonequivalent Groups Design
Pre-Existing groups measured before and after condition, compare changes across groups
Moderate - Can see if change in treatment group was bigger than the natural change in the comparison group
Interrupted Time-Series Design w/ Nonequivalent Groups
Multiple measurements for two groups, before and after conditions introduced
-Compares changes in outcomes across groups
Moderate-to-Strong - Better evidence than pretest/posttest
-Has a comparison group so events that affect both groups (ex; history) can be ruled out because it accounts for general trends/interval variations
Pretest-Posttest w/ Switching Replication
One group receives intervention, the control group waits
-The control group then receives intervention after (staggered introduction)
Strong - Includes a “built-in” replication provides evidence for the effect in two different samples; outside event is unlikely to coincide with both staggered introductions
Switching Replication w/ Treatment Removal
Treatment is removed from the first group at the same time it is added to the second group (when the intervention’s effects are reversible)
-Strong - demonstrating arrival of an effect in two groups and its reversal upon removal
Internal Validity Threats in Non-Equivalent Group Designs
Group Selection Bias
Differential History (History x Group)
Differential Maturation (Maturation x Group)
Differential Attrition (Attrition x Group)
Group Selection Bias
Initial difference between groups in ways that could affect the DV, and confound the IV
-If groups differ on variables that affect DV, any observed effects may reflect initial differences rather than intervention
Differential Maturation (Maturation x Group)
Groups undergo naturally occurring changes or developments at different rates, independently of the intervention
Differential History (History x Group)
An external event affects the groups differently
Differential Attrition (Attrition x Group)
Participants drop out of the study at different rates or for systematically different reasons across conditions, potentially making the remaining groups less comparable
Exchangeability Assumption
Degree to which participants in different conditions are comparable (not systematically different) in ways that could affect the DV
-If we were to exchange participants between conditions, would we expet similar outcomes within each condition
“How Did They End Up There” Test - Are the reasons that determine which condition participants end up in also likely to affect DV independently of intervention?
-If Yes = Potential confounding = less plausible exchangeability
-If No = More plausible exchangeability = more credible causal inference
If Exchangeability is Strong enough
Assignment works AS IF it were random with respect to the DV
Observed differences in DV less likely to be caused by confounding variables
Causal interpretation of results becomes more credible
Factorial Designs
Include more than one IV, each level of one IV is combined with each level of the others to produce all possible combinations
-If a study has two IVs with two levels each, it is a “2×2 factorial design”
Typically has Random Assignment
Between-Subjects Factorial Design
All of the IVs are manipulated between subjects (participants tested in a SINGLE condition)
Mixed Factorial Design
Design which manipulates one IV between subjects and another IV within subjects
Non-Manipulated Independent Variables
IV is measured but NOT manipulated
-Non-manipulated IVs are usually participant variables (gender, self-esteem) and are therefore between-subject factors
-Still have one IV that is manipulated
Results in Factorial Designs
Main Effects
Interaction Effects
Simple Effects
Main Effect
Effect of one IV on the DV, averaging across the levels of the other IV
-One main effect is considered for each IV in the study
Interaction Effect
Effect of one IV depends on the level of the other IV
Types of Interactions:
Spreading Interaction
Cross-Over Interaction
Spreading Interactions
Presence of an effect of one IV at ONE level of the other IV
-Weak effect/no effect of the IV at any other levels of the other IV
Cross-Over Interaction
IV has an effect at both levels, but the effects are in the opposite directions

Simple Effects
Breaking down the interaction to determine precisely what is going on
-Allows researchers to determine the effects of each IV at each level of the other IV
-When interactions are detected, researchers must conduct analyses to determine where that interaction is coming from