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

Top images from around the web for Definition and Formula

• 

  Pearson correlation coefficient - Wikipedia View original

  Is this image relevant?

• 

  Pearson correlation coefficient - Wikipedia View original

  Is this image relevant?

• 

  Correlation - Wikipedia View original

  Is this image relevant?

• 

  Pearson correlation coefficient - Wikipedia View original

  Is this image relevant?

• 

  Pearson correlation coefficient - Wikipedia View original

  Is this image relevant?

1 of 3

Top images from around the web for Definition and Formula

• 

  Pearson correlation coefficient - Wikipedia View original

  Is this image relevant?

• 

  Pearson correlation coefficient - Wikipedia View original

  Is this image relevant?

• 

  Correlation - Wikipedia View original

  Is this image relevant?

• 

  Pearson correlation coefficient - Wikipedia View original

  Is this image relevant?

• 

  Pearson correlation coefficient - Wikipedia View original

  Is this image relevant?

1 of 3

• 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