11.2 Correlation coefficient and its interpretation

Correlation coefficient measures the strength and direction of the relationship between two variables. It's a key tool in understanding how things are connected, ranging from -1 to +1, with 0 meaning no linear relationship.

This concept builds on covariance, providing a standardized measure of association. By calculating and interpreting correlation, we can make predictions, guide research, and inform decisions across various fields, from economics to psychology.

Correlation Coefficient

Definition and Formula

• Correlation coefficient quantifies strength and direction of linear relationship between two continuous variables
• Denoted as r (sample) or ρ (population)
• Dimensionless quantity ranging from -1 to +1
• Formula for Pearson correlation coefficient $r = \frac{\sum[(x - \bar{x})(y - \bar{y})]}{\sqrt{\sum(x - \bar{x})^2 \sum(y - \bar{y})^2}}$
• Population correlation coefficient uses population means (μx and μy) instead of sample means
• Symmetric measure (correlation between X and Y equals correlation between Y and X)
• Invariant under linear transformations of either variable

Properties and Interpretations

• Sign indicates direction of relationship (positive or negative)
• Magnitude represents strength of linear relationship
• Value of 0 suggests no linear relationship (non-linear relationships may still exist)
• Strength categories: 0.00-0.19 (very weak), 0.20-0.39 (weak), 0.40-0.59 (moderate), 0.60-0.79 (strong), 0.80-1.0 (very strong)
• Coefficient of determination (r²) represents proportion of variance in one variable predictable from the other
• Correlation does not imply causation
• Sensitive to outliers and influential points
• Assumes linear relationship (may not accurately represent non-linear relationships)

Calculating Correlation

Data Organization and Preparation

• Organize data into paired observations (x, y) for each subject or item
• Calculate mean (average) of x and y variables separately
• Compute deviations by subtracting mean of x from each x value and mean of y from each y value
  • Example: For data points (2, 3), (4, 5), (6, 7) with means x̄ = 4 and ȳ = 5, deviations are (-2, -2), (0, 0), (2, 2)

Computation Steps

• Multiply x and y deviations for each pair and sum products (numerator of correlation formula)
• Square x and y deviations separately, sum each set of squares, multiply sums, and take square root (denominator)
• Divide numerator by denominator to obtain correlation coefficient
• Verify calculated coefficient falls within -1 to +1 range
  • Example: Using previous data, r = 8 / (√8 * √8) = 1, indicating perfect positive correlation

Interpreting Correlation

Strength and Direction

• Positive values indicate positive relationship (variables increase or decrease together)
  • Example: Height and weight in humans (taller individuals tend to weigh more)
• Negative values indicate negative relationship (one variable increases as other decreases)
  • Example: Temperature and heating costs (higher temperatures lead to lower heating expenses)
• Magnitude closer to -1 or +1 indicates stronger relationship
• Value of 0 suggests no linear relationship
  • Example: Shoe size and intelligence (likely no meaningful correlation)

Practical Implications

• Correlation coefficient helps predict one variable's behavior based on another
• Useful in various fields (economics, psychology, biology)
  • Example: Correlation between study time and test scores to assess effective study habits
• Guides decision-making in research and policy development
  • Example: Correlation between air pollution and respiratory diseases informing environmental policies
• Assists in identifying potential causal relationships for further investigation

Correlation Coefficient Range

Perfect Correlations

• Correlation of +1 indicates perfect positive linear relationship
  • Example: Converting Celsius to Fahrenheit temperatures
• Correlation of -1 indicates perfect negative linear relationship
  • Example: Relationship between price and quantity demanded in perfectly elastic markets
• Perfect correlations rare in real-world data due to natural variability and measurement error

Intermediate Values

• Values between 0 and ±1 indicate varying degrees of linear relationship
• Strength increases as absolute value approaches 1
  • Example: Correlation of 0.7 between exercise frequency and cardiovascular health (strong positive relationship)
  • Example: Correlation of -0.4 between hours of TV watched and academic performance (moderate negative relationship)
• Interpretation depends on context and field of study
  • Example: In social sciences, correlations of 0.3 might be considered meaningful, while in physical sciences, higher correlations may be expected

Limitations and Considerations

• Correlation coefficient sensitive to outliers and influential points
  • Example: A few extreme data points in stock market analysis can skew overall correlation
• Assumes linear relationship (may not accurately represent non-linear relationships)
  • Example: Relationship between age and height in humans (linear in childhood, non-linear in adulthood)
• Restricted range of either variable can affect correlation value
  • Example: Studying correlation between IQ and job performance only for high IQ individuals may underestimate true correlation