Quasi-Experiments
Methodological Toolbox
➔ construct validitytool box
◆ manipulation checks
◆ constants
◆ conceptual/operational definitions
◆ instrument accuracy
➔ external validitytool box
◆ random sampling
◆ ecological validity
◆ experimental realism
➔ statistical validitytool box
◆ sample size
◆ replicability
◆ effect size
◆ significance
➔ internal validitytool box
◆ random assignment
Internal Validity
➔ the3 conditionsof claiming casualty
◆ covariance:as one variable changes, so does the other
◆ temporal precedence:must establish the directionof causality
◆ internal validity:no other explanations exists forthe effect (no confounds)
➔ 3 ways to establish internal validity
◆ comparison groups
● carefully select or create ‘equal’ comparison groups
○ matching or equating in a non-experimental context
◆ temporal precedence
● study the phenomenon longitudinally(over time)
● get a sense of when the IV is introduced and when the DV (post-IV) is
introduced
● measure before the IV, during the IV, and after the IV
◆ rule out alternative explanations
● isolate and control for potential confounds
Review of Terms
➔ IV: a factor that impacts the DV
➔ mediators: explain why IV impacts the DV
, ➔ moderators: explains when the effect occurs (i.e. gender)
➔ intervention: the treatment or the causal agent
◆ the general IV outcome
Experiments
➔ the best way to test causality claims
◆ shows covariance bymanipulating conditionsof theIV
● comparison groups
● allows for clearer comparisons of outcomes of the DV
◆ shows temporal precedence bymanipulating IV beforemeasure the DV
● establishes directionality of the effect
◆ rules out alternative explanations bycontrollingfor them
● randomly assign participants to conditions
○ best way to create unbiased groups for comparison
● holding all factors constant expect for the IV
○ reduces confounds and helps in isolating cause-and-effect
➔ not always possible or ideal
◆ logistically- may not be possible to randomly assignparticipants to conditions
● i.e. does smoking cause cancer?can’t randomly assignparticipants to smoke to
create a condition group
◆ ethically- some questions can’t be ethically addressedwith an experimental study
● i.e. are smartphones ruining our relationships?digginginto people’s personal
relationships and could cause harm to those relationship
◆ generalizability- desire to examine phenomenon ina less controlled, more real-world
context
● i.e. what are the effects of social isolation?can’tmimic social isolation and assume
the effects will be the same
Quasi-Experiments
➔ digs deeper into “why” something occurs
➔ the bridge between correlation studies and true experiments
➔ attempts to establish the three conditions of casualty in ways other than true experiments
◆ covariance: carefully selects comparison groups
◆ temporal precedence: study the phenomenon over time
◆ internal validity: isolate and control for alternativeexplanations
● can only control forknownconfounds
Worst to Best Design
➔ one-group design
◆ introduce a treatment within one group of participants and measure the outcomes
◆ the worst design because doesn’t include comparison group
Methodological Toolbox
➔ construct validitytool box
◆ manipulation checks
◆ constants
◆ conceptual/operational definitions
◆ instrument accuracy
➔ external validitytool box
◆ random sampling
◆ ecological validity
◆ experimental realism
➔ statistical validitytool box
◆ sample size
◆ replicability
◆ effect size
◆ significance
➔ internal validitytool box
◆ random assignment
Internal Validity
➔ the3 conditionsof claiming casualty
◆ covariance:as one variable changes, so does the other
◆ temporal precedence:must establish the directionof causality
◆ internal validity:no other explanations exists forthe effect (no confounds)
➔ 3 ways to establish internal validity
◆ comparison groups
● carefully select or create ‘equal’ comparison groups
○ matching or equating in a non-experimental context
◆ temporal precedence
● study the phenomenon longitudinally(over time)
● get a sense of when the IV is introduced and when the DV (post-IV) is
introduced
● measure before the IV, during the IV, and after the IV
◆ rule out alternative explanations
● isolate and control for potential confounds
Review of Terms
➔ IV: a factor that impacts the DV
➔ mediators: explain why IV impacts the DV
, ➔ moderators: explains when the effect occurs (i.e. gender)
➔ intervention: the treatment or the causal agent
◆ the general IV outcome
Experiments
➔ the best way to test causality claims
◆ shows covariance bymanipulating conditionsof theIV
● comparison groups
● allows for clearer comparisons of outcomes of the DV
◆ shows temporal precedence bymanipulating IV beforemeasure the DV
● establishes directionality of the effect
◆ rules out alternative explanations bycontrollingfor them
● randomly assign participants to conditions
○ best way to create unbiased groups for comparison
● holding all factors constant expect for the IV
○ reduces confounds and helps in isolating cause-and-effect
➔ not always possible or ideal
◆ logistically- may not be possible to randomly assignparticipants to conditions
● i.e. does smoking cause cancer?can’t randomly assignparticipants to smoke to
create a condition group
◆ ethically- some questions can’t be ethically addressedwith an experimental study
● i.e. are smartphones ruining our relationships?digginginto people’s personal
relationships and could cause harm to those relationship
◆ generalizability- desire to examine phenomenon ina less controlled, more real-world
context
● i.e. what are the effects of social isolation?can’tmimic social isolation and assume
the effects will be the same
Quasi-Experiments
➔ digs deeper into “why” something occurs
➔ the bridge between correlation studies and true experiments
➔ attempts to establish the three conditions of casualty in ways other than true experiments
◆ covariance: carefully selects comparison groups
◆ temporal precedence: study the phenomenon over time
◆ internal validity: isolate and control for alternativeexplanations
● can only control forknownconfounds
Worst to Best Design
➔ one-group design
◆ introduce a treatment within one group of participants and measure the outcomes
◆ the worst design because doesn’t include comparison group
