Lecture 4 Notes: Internal Validity and Experimental Design

Internal Validity in Experiments

  • Introduction to Internal Validity

    • Focus of the lecture: Understanding how experiments establish internal validity when other methods do not.

    • Importance of ruling out variables besides the independent variable to confirm causal relationships.

  • Causation and Experimental Design

    • Essential factors for establishing causation in experiments:

    • Covariance: Achieved by having a statistically significant difference in means.

    • Temporal Precedence: Established as the independent variable precedes the dependent variable.

    • Internal Validity: Focus of discussion; ensured through random assignment, which minimizes alternative explanations.

  • Role of Random Assignment

    • Random assignment eliminates alternative explanations and third variables, unlike multiple regression which can only rule out some third variables.

    • Well-designed experiments leave no room for third variables impacting the outcome; they establish a causal link between variables, although the word "prove" should not be used in scientific discussions.

    • Importance of understanding that experiments establish causal links rather than proof of causation.

  • Common Experimental Designs

    • Pharmaceutical Research Example:

    • Participants receive either a real drug or a placebo.

    • Depression scores are measured after the treatment period to determine efficacy of the medication.

    • Ensures that if random assignment is properly conducted and no significant internal validity threats exist, conclusions can reflect a causal relationship.

    • Experimental Control Paradigm:

    • Clinical Psychology Example: Comparing a new therapeutic technique against a waitlist group as a control condition.

    • Cognitive Psychology Example: Examining effects of background noise on attention through a controlled environment.

    • Generally, psychology tends towards comparison approaches rather than strictly control approaches.

  • Comparison vs Control Conditions

    • Comparison Group: One that receives some treatment or intervention but not the same intensity or type as the experimental group.

    • Example: Comparing a new therapy to existing therapies instead of a waitlist or no treatment.

    • At least one comparison or control group is necessary; lacking either makes results uninterpretable.

    • One Sample Experiments: Rare situations where a single group can be tested, typically found in specific statistical conditions, such as one-sample t-tests.

  • Experimental Methods in Psychology

    • Science is defined as a method rather than specific disciplines like chemistry or biology.

    • Example: Acidic vs. basic liquids demonstrated through litmus paper, illustrating how controlled conditions can yield reliable conclusions.

    • Clinical Psychology Example: Assessing the effectiveness of wilderness therapy against antidepressants with randomly assigned samples.

  • Interpreting Results

    • Concerns about interpreting results when only small numbers of participants are used in each condition.

    • Example comparing two individuals receiving different treatments; intuitive doubts about causation despite adherence to statistical principles, highlighting people's tendency to view individual behavior variability differently than chemical/physical experiments.

    • Group Equivalence: Random assignment maintains comparable groups in regards to significant variables.

    • Individuals maintain characteristic variability, but random assignment averages these differences across conditions.

  • Addressing Individual Differences

    • Random assignment isn’t a box to tick; it is an effective method for ensuring equal distribution of participant characteristics.

    • The necessity of having an adequately large sample size ensures that results can be generalized and trustworthy.

  • Random Assignment vs. Random Sampling

    • Random Assignment: Allocates participants to conditions within the experiment to control for internal validity.

    • Random Sampling: Ensures diversity in the sample and external validity of results.

    • Distinction is crucial; random assignment addresses internal validity while random sampling addresses generalizability.

  • Examples Illustrating Random Assignment

    • A memory experiment example comparing studying pictures versus words shows how random assignment negates potential influencing factors related to individual aptitudes for memory.

    • Random assignment enables the researcher to focus solely on the treatment effect and eliminates potential confounders.

  • Conclusion and Next Steps

    • In subsequent lectures, alternatives to random assignment, such as matching and using within-subject designs, will be discussed.