11.2
Coefficient of Correlation
The correlation coefficient, r, developed by Karl Pearson in the early 1900s, is numerical and provides a measure of strength and direction of the lin…
Consider the height and weight of 5 athletes. As the height of athletes increases, their weight also increases. So, height and weight are positively correlated.
The scatter plot of the athlete's weight vs. height shows a linear pattern, which needs to be confirmed using a quantitative measure.
The linear correlation coefficient, denoted by r, provides a quantitative measure of the strength of such a linear correlation between two variables.
For such a dataset with n scatter points whose x and y values are known, r can be calculated.
The value of r always lies between -1 and +1. The higher the modulus of r, the stronger the correlation between the variables.
If the value of x or y is swapped, or one of the variables is converted to a different scale, the value of r is not affected.
The coefficient of correlation is strongly affected by outliers. Hence, if such data points are known to be errors, they can be removed to improve the accuracy of the value of r.
View the full transcript and gain access to JoVE Core videos
Q1: What does a correlation coefficient value tell you about the relationship between two variables?
The correlation coefficient r measures the strength and direction of a linear relationship between two variables, with values always between -1 and +1. Values close to -1 or +1 indicate stronger linear relationships, while r = 0 suggests no linear correlation. The sign of r indicates direction: positive values mean variables move together, negative values mean they move oppositely.
Q2: How does the correlation coefficient differ between positive and negative correlations?
A positive correlation coefficient indicates that as one variable increases, the other tends to increase as well, like height and weight in athletes. A negative correlation means that as one variable increases, the other tends to decrease. The sign of the correlation coefficient matches the slope of the best-fit line through the data.
Q3: What happens to the correlation coefficient when data is transformed or scaled?
The correlation coefficient r remains unaffected if the x or y values are swapped or if one variable is converted to a different scale. This property makes r a robust measure for comparing linear relationships across datasets measured in different units or orders.
Q4: Why are outliers problematic when calculating the correlation coefficient?
The correlation coefficient is strongly affected by outliers, which can distort the measured strength of the linear relationship. If outliers are known to be errors, removing them improves the accuracy of the r value and provides a more reliable measure of the true linear association between variables.
Q5: What does it mean when the correlation coefficient equals exactly 1 or -1?
When r = 1, there is a perfect positive correlation, meaning all data points lie exactly on a straight line with positive slope. When r = -1, there is a perfect negative correlation with all points on a line with negative slope. In real-world datasets, these perfect values rarely occur.
Q6: How can you visually confirm a correlation coefficient value?
A scatter plot should always accompany the correlation coefficient calculation to visually confirm the linear relationship. Data exhibiting curved or horizontal patterns may have a correlation of zero despite appearing related, so viewing the scatterplot prevents misinterpretation of the numerical r value.
Q7: Who developed the correlation coefficient and what makes it useful for statistical analysis?
Karl Pearson developed the correlation coefficient in the early 1900s as a numerical measure of linear association strength and direction. It provides a standardized, scale-independent way to quantify relationships between variables, making it essential for comparing correlations across different datasets and research contexts.