Research Control: Validity, Variables & Design

• Control is essential in research to eliminate potential threats to the study's validity.
• Extraneous variables can confound results if not controlled.
• Confounding variables co-vary with the independent variable, leading to confusion.

Achieving Control

• Random assignment equates participant groups, controlling for both known and unknown effects.
• Matching controls for known effects when random assignment isn't possible.
• Manipulation checks assess the effectiveness of the independent variable manipulation.
• Objective measurement instruments enhance control.
• Statistical control methods improve design analysis and statistical power (e.g., increasing sample size).
• Effect size measures the strength of the relationship between variables while power indicates the ability to detect a relationship, if it's present.

Random Assignment

• An equal chance of being assigned to any group or condition is given to every sample member.
• Experimental designs rely on this to equate groups.

Matching

• Variables or characteristics of interest are matched among participants.
• It aims to create equivalent groups without random assignment.
• This only controls for known effects.

Random Sampling

• Should be used in all research design.
• Each population member has an equal chance of being selected.
• It's a process for choosing a representative sample.

Reducing Experimenter Bias

• Double-blind procedures keep both researchers and participants unaware.
• Single-blind procedures only keep the participant uninformed.
• Include multiple researchers in the study design.
• Standardize researcher behavior during interactions with participants.
• Deception may be used but needs debriefing, although debriefing doesn't directly control bias.

Between-Subjects Design

• Different participants are placed at different independent variable levels.
• Participants experience only one condition.
  • Advantage: Minimizes testing effects.
  • Disadvantage: Requires more participants, less assurance of equivalency.

Within-Subjects Design

• Each participant experiences all independent variable levels.
  • Advantage: Ensures equivalence, requires fewer participants.
  • Disadvantage: Repeated testing effects, irreversibility of treatment effects, dependability of treatment effects (order effects and carryover effects).

Strategies to Enhance Control in Within-Subjects Design

• Practice effects are controlled by increasing time intervals.
• Counterbalancing reverses, uses intrasubject or intersubject methods, and can be complete or incomplete.
• Carryover effects are contaminations from one condition to another.
• Order effects result from the position of the condition alone.

Factorial Designs

• Simple factorial designs involve one independent variable with at least two levels.
• Complex factorial designs involve multiple independent variables.
• Mixed factorial designs use both within-subjects and between-subjects designs.

ANOVA

• Used to simultaneously compare two or more means
• Used to study the joint effect of multiple independent variables.

Why Experiments Support Claims

• Cause equals A and Effect equals B.
  • Continuity: The cause is related to the effect (A is related to B).
  • Temporal Precedence: The cause comes before the effect (A before B).
  • Constant Conjunction: The cause is present when the effect is obtained (A present when B is obtained).

Causal Inferences

• Experimental designs permit causal inferences.
• Causation implies the ability to make statements about cause-effect relationships.

Main Effects

• Determines if there's an overall significant difference.
• Examines if there's an overall effect of one independent variable on a dependent variable.

Interactions

• Independent variables interact when they act upon one another.
• Interactions are always moderators.
• A 2x2 design has 2 main effects and 1 interaction.
• A 2x2x3 design has 3 main effects and 4 interactions.
• A moderator influences the strength/direction of the relationship between variables.
• A mediator is the cause of the effect.

Inferential Statistical Tests

• F is the test statistic.
• P-value is alpha, determining statistical significance.
• N^2 is the effect size, indicating effect magnitude.
• Parallel lines on ANOVA plots suggest a non-significant interaction.