Experimental Designs and Causal Claims

Research Goals and Claims

• Describe behavior: Make a claim about frequency.
• Example: How prevalent is depression?
• Predict behavior: Make a claim about associations.
• Example: Is depression related to spending time alone?
• Explain behavior: Make a causal claim.
• Example: Does social isolation make people depressed?

Elements of an Experiment

• Independent Variable (IV): The variable that the experimenter manipulates.
• e.g., Treatment Group, Control Group
• Levels or Conditions: Must have at least two levels.
  • Example Levels: Low intensity, Medium intensity, High intensity.
• Dependent Variable (DV): The variable that is measured.

Experimental Design Elements

• Experimental Control: Holding all variables constant except for the IV to ensure a fair test.
• Random Assignment: Participants are randomly assigned to any given condition, ensuring equal probability among conditions.

Criteria for Making Causal Claims

1. Covariance: Cause and effect must co-occur.
2. Temporal Precedence: The cause must precede the effect.
3. Internal Validity: The experiment must control for third variables.

Establishing Covariance

• Use a comparison group to address the question, "Compared to what?"

Establishing Temporal Precedence

• The manipulation of the IV must occur before measuring the DV.

Establishing Internal Validity

• Confounds: A variable that may systematically vary with the levels of the IV and affect the DV.
• Noise Variables: Variables that do not systematically vary with the IV but can still affect the DV.

Selection Effects

• Occurs when participant types differ systematically between conditions, often resulting from self-selection.

Random Assignment & Matched Groups Design

• Random assignment turns potential confounds into noise variables.
• Matched Groups Design: Pair participants based on a variable (e.g., GPA) and randomly assign them to groups.

Types of Experimental Designs

• Independent-groups designs:
• Posttest only
• Pretest & posttest
• Matched-groups design
• Within-groups designs:
• Repeated measures
• Concurrent measures

Evaluating Causal Claims

1. Internal Validity: Measures if the IV truly affects the DV.
2. Statistical Validity: Considers sample size effect and power.
3. External Validity: Questions generalizability to real-life scenarios.
4. Construct Validity: Checks if the manipulation effectively changes the IV and accurately measures the DV.

Example Study - Latané and Darley (1969)

• Aim: Study the bystander effect.
• Method: 120 male participants completed a survey. Hear a recording of a female experimenter supposedly falling and needing help.
• IV: Survey condition (alone or with a confederate)
• DV: Whether participants helped or not.
• Findings: 70% helped when alone; only 7% helped with a passive confederate.

Evaluation Questions

• Identify IV, DV, and controls.
• Type of study: Posttest only independent groups design.
• Evaluating based on four validities:
• Construct Validity: How well the IV and DV are measured.
• External Validity: Generalizability beyond the sample.
• Statistical Validity: Includes sample size and power analysis.
• Internal Validity: Control of confounding variables.