Please enable JavaScript.
Coggle requires JavaScript to display documents.
PS906: Non-Randomized Designs - Coggle Diagram
PS906: Non-Randomized Designs
Types
Quasi-Experimental Design
Where an experimental procedure is applied
But NOT all extraneous variables can be controlled
Limited Causal Inferences
Cause and Effect covary
Cause precede Effect
Challenge: Confounding variables - difficult to control
Implications
Internal Validity
Causal Inference
Correlational Studies
Case Studies
Survey Research
Types of Quasi-Experimental Design
Non-equivalent comparison group design
most common
includes different groups BUT participants are not randomly assigned
therefore groups are not equivalent on all (confounding) variables
uncontrolled variables operate as rival hypotheses to explain experimental outcome
Ruling out THREATS (to internal validity)
Statistical Control
Measure variables for which one might find differences between groups in the pre-test
Include variable(s) as covariates
Beware of measurement error!
Matching
Match groups as closely as posisble on variables that might pose rival explanations
Time-Series Design
used when hard to find equivalent group of participants as control group
collects data over time to observe changes in a variable or phenomenon
study trends, patterns & relationships over time
Interrupted
Involves measuring a dependent variable over time, introducing an intervention, and then continuing to measure the dependent variable
e.g. Measuring the number of traffic accidents at an intersection before and after installing a new traffic light.
e.g. Measure the number of workplace accidents before and after implementing the new safety policy
challenge: possible due to history (as there's no control group) - any events that happen in addition to the treatment
Regression Discontinuity Design
useful when wanting to investigate the efficacy of a treatment but cannot randomly assign participants to groups
threats: differential history - events that affect only one group, not the other
Assigns participants to treatment based on a cutoff score on a continuous variable
e.g. Assigning students to a remedial program based on their standardized test scores, with the cutoff score determining eligibility
e.g. Assign employees to promotions based on a performance cutoff score, and compare turnover rates between those who were promoted and those who were not
Single Case Design
research methodology that focuses on studying a single individual, group, or organization in depth
useful for exploring unique cases, testing theories, and understanding complex phenomena
ABA Design
withdrawal of treatment
(back to baseline) - reversal
variation
ABAB
further strengthen causal relationship
Example
ABAB Design: Improving Employee Productivity
Scenario: A manager wants to increase employee productivity in a call center.
Baseline (A): The average number of calls handled per employee per hour is measured for two weeks.
Intervention (B): A new incentive program is implemented, offering bonuses for high-performing employees. The average number of calls handled is measured for two weeks.
Baseline (A): The incentive program is temporarily removed, and the average number of calls handled is measured again for two weeks.
Intervention (B): The incentive program is reintroduced, and the average number of calls handled is measured for another two weeks.
Multiple Baseline Design
Example 1
Scenario: A behavior analyst wants to assess the impact of a token economy system on three different problem behaviors in a child with autism.
Baseline 1: The frequency of tantrums is measured for two weeks.
Baseline 2: The frequency of self-injury is measured for two weeks.
Baseline 3: The frequency of property destruction is measured for two weeks.
Intervention: The token economy system is introduced to reduce all three behaviors. However, the intervention is implemented for each behavior at a different time.
Example 2
Multiple Baseline Design: Reducing Employee Absenteeism
Scenario: A company wants to reduce employee absenteeism across three different departments.
Baseline 1: The absentee rate in the sales department is measured for two months.
Baseline 2: The absentee rate in the customer service department is measured for two months.
Baseline 3: The absentee rate in the production department is measured for two months.
Intervention: A new wellness program is introduced, offering health screenings, fitness classes, and stress management workshops. However, the program is implemented in each department at a different time.
Reversal Design
repeatedly introduced and withdrawn to demonstrate a clear cause-and-effect relationship
Example
Reversal Design: Improving Employee Morale
Scenario: A manager wants to evaluate the effectiveness of a new employee recognition program on morale.
Baseline (A): Employee satisfaction surveys are conducted every month for three months.
Intervention (B): The new recognition program is implemented, featuring public praise, awards, and social events. Surveys continue for three months.
Reversal (A): The recognition program is temporarily stopped, and surveys continue for three months.
Intervention (B): The recognition program is reintroduced, and surveys continue for another three months.