Module 0.4 Correlation and Experimentation Study Notes

Overview of Research Methods and Learning Targets

  • Psychologists utilize diverse non-experimental and experimental research methods to describe, predict, and explain human thinking, feeling, and behavior.

  • Correlational research is a non-experimental research design used to describe and quantify the strength and direction of the relationship between two or more variables.

  • Experiments involve active manipulation of one or more factors of interest while holding other variables constant to establish explicit cause-and-effect connections.

  • Module Learning Targets:

    • Explain what it means when two factors correlate, and distinguish between positive and negative correlations.

    • Explain illusory correlations and the phenomenon of regression toward the mean.

    • Describe the defining characteristics of experimentation that make it possible to isolate cause and effect.

Nature and Properties of Correlations

  • Definition of Correlation: A statistical measure indicating the extent to which two variables vary together, and consequently how well either variable predicts the other.

  • Observational Foundation: Describing behavior through naturalistic observations or surveys serves as the initial step toward prediction by identifying traits or behaviors that coincide.

  • Variable: Any factor, attribute, or characteristic that can exist in differing amounts or types (e.g., personality test scores in identical twins, intelligence test scores, career achievement, aptitude test scores, or academic success).

  • Positive Correlation: Describes a relationship where two sets of scores tend to increase or decrease together in the same direction (e.g., height and weight).

  • Negative Correlation: Describes an inverse relationship where one set of scores increases as the other set decreases (e.g., a person's height and the distance from the top of their head to the ceiling).

  • Direction vs. Value Judgment: Calling a correlation "negative" denotes purely its inverse direction; it carries no value judgment, does not imply a "bad" outcome, and says nothing about the strength of the association.

Correlation Coefficients and Scatterplots

  • Correlation Coefficient (rr): A statistical index that quantifies the direction and strength of the relationship between two variables, ranging from −1.00-1.00 to +1.00+1.00:

    • Perfect Positive Correlation (r=+1.00r = +1.00): Scores for one variable increase in exact direct proportion to scores for another variable.

    • No Correlation (r=0.00r = 0.00): No linear relationship exists between the two variables.

    • Perfect Negative Correlation (r=−1.00r = -1.00): Scores for one variable decrease in exact inverse proportion as scores for another variable increase.

  • Scatterplot: A graphed cluster of dots, each representing the values of two variables for a single case or observation.

Scatterplots showing patterns of correlation
  • Real-World Frequency: Perfect correlations (+1.00+1.00 or −1.00-1.00) rarely occur in natural human behavior or psychological measurements.

Fear and Disgust Animal Study

  • Statistical Illumination: Statistical analysis brings to light patterns and relationships in empirical data that casual observation easily misses.

  • Study Parameters (Polák et al., 2020): 22912291 Czech and Slovakian volunteers rated their levels of fear and disgust on a 11 to 77 scale in response to 2424 different animals.

  • Empirical Dataset (Table 0.4-1 - Average Fear and Disgust Ratings):

    • Ant: Average Fear = 2.122.12, Average Disgust = 2.262.26

    • Bat: Average Fear = 2.112.11, Average Disgust = 2.012.01

    • Bull: Average Fear = 3.843.84, Average Disgust = 1.621.62

    • Cat: Average Fear = 1.241.24, Average Disgust = 1.171.17

    • Cockroach: Average Fear = 3.103.10, Average Disgust = 4.164.16

    • Dog: Average Fear = 2.252.25, Average Disgust = 1.201.20

    • Fish: Average Fear = 1.151.15, Average Disgust = 1.381.38

    • Frog: Average Fear = 1.841.84, Average Disgust = 2.482.48

    • Grass snake: Average Fear = 3.323.32, Average Disgust = 2.472.47

    • Horse: Average Fear = 1.821.82, Average Disgust = 1.111.11

    • Lizard: Average Fear = 1.461.46, Average Disgust = 1.461.46

    • Louse: Average Fear = 3.583.58, Average Disgust = 4.834.83

    • Maggot: Average Fear = 2.902.90, Average Disgust = 4.494.49

    • Mouse: Average Fear = 1.621.62, Average Disgust = 1.781.78

    • Panda: Average Fear = 1.571.57, Average Disgust = 1.171.17

    • Pigeon: Average Fear = 1.481.48, Average Disgust = 2.012.01

    • Rat: Average Fear = 2.112.11, Average Disgust = 2.252.25

    • Rooster: Average Fear = 1.781.78, Average Disgust = 1.341.34

    • Roundworm: Average Fear = 3.493.49, Average Disgust = 4.794.79

    • Snail: Average Fear = 1.151.15, Average Disgust = 1.691.69

    • Spider: Average Fear = 4.394.39, Average Disgust = 4.474.47

    • Tapeworm: Average Fear = 3.603.60, Average Disgust = 4.834.83

    • Viper: Average Fear = 4.344.34, Average Disgust = 2.832.83

    • Wasp: Average Fear = 3.423.42, Average Disgust = 2.842.84

  • Scatterplot Visualization: Although scanning raw tabular data makes the relationship obscure, plotting these data points on a scatterplot reveals a distinct positive correlation of r=+0.72r = +0.72.

Scatterplot for fear and disgust felt toward 24 animals

Applied Examples of Correlation

  • Sleep and Academic Performance (Okano et al., 2019): Amount of sleep in college students correlates positively with academic performance.

  • Social Media Use and Mental Health Risk (Kelly et al., 2019; Twenge & Campbell, 2019): Hours spent on social media among adolescent girls correlate positively with risks for depression and suicidal thoughts.

  • Breast-Feeding and Cognitive Achievement (Horwood & Ferguson, 1998): Duration of infant breast-feeding correlates positively with later academic achievement.

  • Vegetable Consumption and Cognitive Decline (Morris et al., 2018): Leafy vegetable consumption in older adults correlates negatively with cognitive decline over the following 55 years (higher vegetable intake correlates with reduced mental decline).

  • Systematic Statistics vs. Casual Observation (Twiss et al., 1989): Gender discrimination in compensation and promotion becomes clear when examining aggregate statistical summaries across job level, seniority, performance, gender, and salary, whereas case-by-case observation obscures bias because isolated individual cases capture subjective attention.

Correlation and Causation

  • Fundamental Rule: Correlation does not equal causation. A correlation indicates that a relationship exists and enables prediction, but it does not prove that one variable causes changes in the other.

  • Directionality Problem: Correlational data cannot reveal which factor is the cause and which is the effect:

    • Example: Teen social media use correlates with depression. Social media use might cause depression, or preexisting depression might cause increased social media use.

  • Third Variable Problem: An unmeasured third factor may produce the observed association between two variables:

    • Smoking and Mental Illness: Smoking correlates with mental illness. Smoking could cause mental illness, mental illness could lead to smoking, or a stressful home environment (third variable) could trigger both.

    • Intimacy and Well-Being: Sexual restraint correlates with relationship stability and mental health. Sexual restraint could enhance relationships, depression could increase casual hookups, or lower trait impulsivity (third variable) could drive both restraint and better mental health.

    • Self-Esteem and Depression: Low self-esteem correlates negatively with depression. Low self-esteem could cause depression, depression could lower self-esteem, or distressing life events / biological predispositions (third variable) could cause both.

    • Parental Affection and Risk Behavior: A survey of over 1200012000 adolescents revealed that feeling loved by parents correlated with reduced rates of early sexual activity, smoking, drug/alcohol abuse, and violence.

    • Marriage Length and Balding: Length of marriage correlates positively with male hair loss. Marriage does not cause hair loss; advancing age is the underlying third variable.

Illusory Correlations and Regression Toward the Mean

  • Illusory Correlation: The perception of a relationship between two variables where none exists, or the perception of a stronger relationship than actually exists.

    • Cognitive Mechanism: Driven by confirmation bias—when holding a belief about a relationship, individuals notice and recall confirming instances (e.g., dreams forecasting events) while ignoring disconfirming instances.

    • Illusion of Control: Illusory correlations foster the belief that personal actions influence chance outcomes (e.g., gamblers throwing dice gently for low numbers and hard for high numbers).

  • Regression Toward the Mean: The statistical tendency for extreme or unusual scores or events to fall back (regress) toward their average value over time.

    • Underlying Cause: Extraordinary occurrences are driven by temporary, chance combinations of factors (e.g., test difficulty, sleep quality, weather) that are unlikely to repeat in identical fashion.

    • Test Scores: An unusually low test score caused by unfortunate conditions is typically followed by a score closer to the student's personal average.

    • Athletic Performance: An unusually poor game by a sports team is usually followed by an improved performance returning to average.

    • Superstitious Interpretations in Coaching (Tversky & Kahneman, 1974):

    • A coach who scolds a team following an extraordinarily bad performance falsely concludes that criticism caused the subsequent return to normal performance.

    • A coach who praises a team after an extraordinary performance falsely concludes that praise caused the subsequent drop back toward average.

    • Key Insight: Psychologist Daniel Kahneman (1985) noted, "Once you become sensitized to it, you see regression everywhere." When fluctuating behavior returns to normal, extraordinary explanations are usually wrong; regression toward the mean is the primary mechanism.

Fundamentals of Experimentation

  • Definition of Experiment: A research method in which an investigator manipulates one or more independent variables to observe the effect on a dependent variable while controlling other potential confounding variables.

  • Primary Advantage: Experimentation isolates cause and effect, providing causal evidence that non-experimental correlational methods cannot establish.

  • Philosophical Context: Roman poet Virgil stated, "Happy are they who have been able to perceive the causes of things."

  • Empirical Context on Teen Mental Health:

    • Phenomenon 1: Smartphone and social media adoption accelerated rapidly starting in 20102010

    • Phenomenon 2: Rates of depression, anxiety, self-harm, and suicide among Canadian, American, and British adolescent girls increased substantially during the same period (Mercado et al., 2017; Morgan, 2017; Statistics Canada, 2016).

    • Longitudinal Predictive Power: In 77 out of 99 longitudinal studies, adolescent social media use predicted future mental health issues (Haidt & Twenge, 2019; Zhou et al., 2020).

  • Facebook Deactivation Experiment (Allcott et al., 2020):

    • Sample: Nearly 17001700 participants agreed to deactivate their Facebook accounts for 44 weeks.

    • Experimental vs. Control Outcomes: Participants randomly assigned to deactivate Facebook spent more time watching TV and socializing in person; they reported lower depression, higher happiness, greater life satisfaction, and reduced post-experiment Facebook usage relative to the control group.

    • Scientific Consensus: Most researchers agree unlimited adolescent social media use poses a modest risk to mental health.

Experimentation design diagram

Random Assignment and Grouping

  • Experimental Group: The group in an experiment exposed to the treatment condition (the specific manipulation of the independent variable, such as deactivating social media).

  • Control Group: The comparison group in an experiment that does not receive the experimental treatment; serves as a baseline for comparison.

  • Random Assignment: Assigning participants to experimental and control groups by chance (e.g., using a random numbers table or coin flip), minimizing preexisting differences between groups.

    • Function: Equalizes groups across background variables (e.g., age, attitudes, temperament, or physical traits like ear wiggling—if 1/31/3 of participants can wiggle their ears, approximately 1/31/3 in each group will have that ability).

  • Difference Between Random Sampling and Random Assignment:

    • Random Sampling: Selects a representative sample from a broader population for survey or observational studies.

    • Random Assignment: Distributes participants into experimental and control conditions within an experiment to eliminate confounding factors.

Experimental Controls, Placebo Effect, and Blinding Procedures

  • Evaluation of Therapies: Misleading testimonies frequently arise when illnesses naturally subside over time (e.g., taking zinc tablets 33 days into a common cold, or historical 1700s1700\text{s} bloodletting where patient survival was attributed to the treatment and death was attributed to disease severity).

  • Inert Placebo: A fake treatment or inactive substance (e.g., a pill containing no active medicinal ingredient) given to control conditions.

  • Placebo Effect: Experimental results caused by expectations alone; any effect on behavior or physiology caused by the administration of an inert substance or condition.

    • Demonstrations: Reduces pain, depression, anxiety, and auditory hallucinations in schizophrenia (Dollfus et al., 2016; Kirsch, 2010); increases running speeds in athletes believing they took performance drugs (McClung & Collins, 2007); enhances alertness in decaf coffee drinkers believing they consumed caffeine (Dawkins et al., 2011).

    • Cost Bias: A fake pill costing $2.50\$2.50 produces stronger therapeutic placebo effects than a fake pill costing $0.10\$0.10 (Waber et al., 2008).

  • Single-Blind Procedure: An experimental procedure where research participants are unaware (blind) of whether they are receiving the treatment or a placebo, controlling for social desirability bias.

  • Double-Blind Procedure: An experimental procedure where both the research participants and the research staff administering treatment and collecting data are unaware (blind) of group assignments.

    • Function: Simultaneously eliminates participant expectations (placebo effect) and experimenter bias (unintentional influence by researchers to confirm hypotheses).

Independent, Dependent, and Confounding Variables

  • Independent Variable (IV): The factor that is manipulated in an experiment; the variable whose effect is being evaluated.

  • Dependent Variable (DV): The outcome factor that is measured; the variable that may change in response to manipulations of the independent variable.

  • Operational Definitions: Precise statements of the exact procedures and measurements used in a research study to define the independent and dependent variables, enabling exact replication by other researchers.

  • Confounding Variables: Factors other than the independent variable that might influence experiment results (e.g., participant temperament, environmental differences). Random assignment controls for potential confounding variables.

  • Quiz Testing Experiment (Victor Benassi et al. / Lee et al., 2014):

    • Manipulation (IV): Study procedure during frequent quizzes (reading answer key versus actively generating answers/self-testing).

    • Outcome (DV): Final examination performance measured weeks later.

    • Results: Students scored 75%75\% correct on self-tested material compared to 51%51\% correct on restudied material.

  • Rental Housing Discrimination Experiment (Carpusor & Loges, 2006):

    • Design: Identically worded email inquiries sent to 11151115 Los Angeles-area landlords.

    • Independent Variable: Ethnic connotation of sender's name ("Patrick McDougall", "Said Al-Rahman", "Tyrell Jackson").

    • Dependent Variable: Percentage of positive landlord responses offering apartment viewings.

    • Results: "Patrick McDougall" received 89%89\% positive replies, "Said Al-Rahman" received 66%66\%, and "Tyrell Jackson" received 56%56\%.

Validity and Scientific Principles

  • Validity: The extent to which a test or experiment measures or predicts what it is intended to measure or test.

    • Applied Check: In the rental housing experiment, validity assesses whether email inquiries measured the specific effect of perceived ethnicity on landlord responses.

    • Distinction from Reliability: Reliability denotes the consistency of measurement and ability to replicate findings across trials.

  • Societal Context: President Barack Obama stated in his June 26, 2015 eulogy for Clementa Pinckney: "[We must guard] against not just racial slurs, but… against the subtle impulse to call Johnny back for a job interview, but not Jamal."

Concept Assessment and Practice Solutions

  • Matching Definitions:

    • Double-blind procedure: Controls for the placebo effect; neither researchers nor participants know who receives the real treatment.

    • Dependent variable: The outcome that is measured in an experiment; the variable that may change when the independent variable is manipulated.

    • Random assignment: Minimizes preexisting differences between experimental and control groups.

  • AP Practice Multiple Choice Questions & Analysis:

    1. Example of Negative Correlation: People who spend more time exercising tend to experience less depression. (As exercise time increases, depression ratings decrease).

    2. Room Temperature and Test Performance Experiment (Independent Variable): The temperature of the room.

    3. Income Levels and Total Sleep Hours (Lower income levels report fewer hours of sleep): Income and sleep levels are positively correlated.

    4. Strongest Relationship Correlation Coefficient (+0.75+0.75, +1.3+1.3, −0.85-0.85, −0.05-0.05): −0.85-0.85 (represented by the highest absolute value; +1.3+1.3 is invalid as correlation coefficients are bounded between −1.00-1.00 and +1.00+1.00).

    5. Primary Purpose of Random Assignment: Reduce potential confounding variables.

    6. Research Design with Uninformed Participants and Staff: Double-blind study.

    7. Purpose of a Control Group in Experimental Design: Allows researchers to determine a cause-and-effect relation between the independent variable and dependent variable.