Correlational Research Notes

Correlational Research: Key Concepts

• A correlation is a relation between two variables; it describes how two things vary together, not why. For example, height and weight vary across people and tend to move together in a general way.
• The two variables can be anything that can be measured and related; there are always individual differences (e.g., tall people can be thin or fat, short people can be heavy or light).
• In psychology, correlations illustrate how one variable tends to vary with another, not causation.

Types of Correlations: Positive and Negative

• Positive correlation (direction): as one variable increases, the other tends to increase as well.
  • Examples:
  • IQ scores and GPA: higher IQ tends to be associated with higher GPA.
  • Motivation (hours studied) and GPA: more study tends to go with higher GPA.
  • Shyness at age 1 and shyness at age 16: more shyness early tends to relate to more shyness later.
  • Note: A positive correlation does not guarantee that increases in one cause increases in the other; it indicates a general pattern.
• Negative correlation (direction): as one variable increases, the other tends to decrease.
  • Examples:
  • Smoking amount and lifespan: more smoking tends to be associated with shorter life expectancy.
  • Episodes of mental illness and income: more mental illness tends to be associated with lower income (though wealthy individuals can still experience mental illness; wealth may enable better treatment).
  • Important: These are general tendencies, not universal rules; there can be exceptions.

Strength of a Correlation vs Type

• Type refers to direction: positive or negative (not to strength).
• Strength refers to how strong the relationship is.
• The strength is quantified by the correlation coefficient, denoted by $r$ (sometimes referred to as R in casual talk).
• The magnitude ranges roughly from 0 to 1 (0 means no relationship; 1 means a perfect relationship). The sign indicates direction.
• In practice, the lecturer treats the magnitude as the strength, with the sign indicating the type.
• Key points:
  • A stronger correlation has a larger magnitude |$r$| (closer to 1 or -1).
  • The sign of $r$ indicates positive ( + ) vs negative ( - ) direction.
  • Positive and negative do not indicate how strong the relationship is.

Interpreting the Correlation Coefficient $r$

• Range and interpretation:
  • Magnitude: (|r|) lies between 0 and 1.
  • Sign: if r > 0, the relationship is positive; if r < 0, the relationship is negative.
  • All else equal, larger |$r$| means a stronger association; the sign tells direction.
• Examples from the lecture:
  • IQ scores and GPA: $r \approx 0.50$ (moderate-to-strong positive relationship).
  • Motivation and GPA: $r \approx 0.60$ (stronger than IQ-GPA; still positive).
  • Smoking and lifespan: $r \approx -0.70\text{ to }-0.80$ (strong negative relationship).
• Rule of thumb shown in the lecture: a correlation around $0.3$ or higher is considered statistically significant in this context.

R and R^2: What They Tell You

• The correlation coefficient is symbolized by $r$ and is a measure of the strength and direction of a linear relationship between two variables.
• The square of the correlation, $r^2$, represents the proportion of variance in one variable that is explained by the other variable (in the simple linear model sense).
• Example with smoking and lifespan:
  • If $r = -0.70$, then $r^2 = (-0.70)^2 = 0.49$, i.e., about 49% of the variance in lifespan is accounted for by variation in smoking (in the sample context; this is a rough interpretation).
  • The lecture notes mention this as about 50%, illustrating the idea that a substantial portion of variability in one variable can be explained by the other when the correlation is strong.
• Important distinction:
  • The sign of $r$ indicates direction (positive vs negative).
  • The magnitude |$r$| indicates strength of the relationship.

Examples in Context: Interpreting Specific Correlations

• IQ and GPA: $r \approx 0.50$ → moderate-to-strong positive correlation; higher IQ tends to go with higher GPA.
• Motivation (hours studied) and GPA: $r \approx 0.60$ → stronger positive correlation than IQ-GPA; greater study time tends to relate to higher GPA.
• Smoking and lifespan: $r \approx -0.70 \text{ to } -0.80$ → strong negative correlation; more smoking tends to relate to shorter life expectancy.

Strengths and Limitations of Correlational Research

• Strengths:
  • Can determine whether two variables are related and how they relate (positive/negative; how strong).
  • Allows examination of relationships without manipulating variables; relatively easy to measure variables.
• Limitations (the big caveat):
  • Correlational studies describe relationships but do not explain why the relationship exists.
  • Correlation does not imply causation (two variables can be related without one causing the other).
  • Possibility of a third variable (a confound) that influences both variables and creates the observed relationship.
  • The relationship could be due to reverse causation or bidirectional influence.
• The classic warning: “Correlation does not imply causation; only co-occurrence.”

The Media Violence Example: Why Correlation Is Not Causation

• Measured variables:
  • Exposure to media violence (e.g., violent movies/sites/music).
  • Actual violent behavior.
• Observed relationship: a positive correlation around $r \approx 0.30$ (a modest but significant relationship).
• Possible interpretations (all plausible, given correlational data):
  • Exposure to media violence causes violent behavior (causal hypothesis).
  • People who are prone to violence are more drawn to media violence (reverse causation).
  • A third variable explains both: e.g., a home environment with high levels of violence influences both exposure to media violence and aggressive behavior (a common root).
• The key takeaway: even with a statistically significant correlation, you cannot determine the causal direction or mechanism from correlational data alone.
• Visual representation concept:
  • A third variable (common root) can influence both variables, creating an observed association.

Testing Causation: The Role of Experiments

• To determine whether one variable causes another, you need an experiment.
• Experimental designs control for confounding variables and use random assignment to infer causality.
• The lecturer emphasizes that experiments are the scientifically valid way to test for cause-and-effect relationships between variables and that this topic will be covered in class.

Practical Takeaways and Connections

• Use correlational analysis to identify relationships and quantify their direction and strength, but be cautious about interpreting causation.
• Always consider potential third variables or confounds that could explain observed correlations.
• When interpreting r and r^2, remember:
  • Direction is given by the sign of r.
  • Strength is given by |r|; larger magnitude means a stronger association.
  • r^2 indicates the proportion of variance in one variable explained by the other in the sample.
• Real-world relevance: in education, health, and psychology, correlations guide hypotheses and further research, including whether to pursue experimental studies to test causality.

Quick Reference Formulas

• Correlation coefficient: $r$ with sign indicating direction and magnitude indicating strength; $-1 \le r \le 1$.
• Proportion of variance explained: $r^2$, where \n - Example: if $r = -0.70$, then $r^2 = 0.49$ (≈ 49% of variance explained).
• Sign interpretation: a plus sign indicates a positive correlation; a minus sign indicates a negative correlation.
• Threshold for significance (as stated in lecture): $|r| \ge 0.3$ indicates a statistically significant relationship in the given context.

Key Takeaway

• Correlational research reveals whether variables are related, the direction of the relationship, and its strength, but it does not prove causation. Experiments are needed to establish causal relations and rule out confounding variables or reverse causation.