Pearson 相关分析：如何正确衡量线性关系

What Is Pearson Correlation?

Pearson correlation measures the strength and direction of a linear relationship between two continuous variables. It answers questions such as whether study time tends to increase with exam scores or whether blood pressure tends to rise with body mass index.

The correlation coefficient is written as r, and it ranges from -1 to +1. Values near +1 indicate a strong positive linear relationship, values near -1 indicate a strong negative linear relationship, and values near 0 suggest little or no linear pattern.

The Formula and What It Represents

Pearson correlation is based on covariance scaled by the standard deviations of the two variables:

r = covariance(X, Y) / (sd(X) x sd(Y))

Because of this standardization, the statistic is unit-free. A change from hours to minutes or from kilograms to pounds will not change the correlation value.

When Is Pearson Correlation Appropriate?

• Both variables are continuous.
• You are interested in a linear relationship.
• The observations are paired, meaning each row contains one value for each variable from the same unit.

Main Assumptions

• The relationship is approximately linear.
• Extreme outliers are absent or handled carefully.
• The observations are independent.
• For formal hypothesis testing, the joint distribution should not be severely distorted.

Step-by-Step Method

1. Plot the data with a scatterplot.
2. Check whether the pattern looks roughly linear.
3. Look for unusual outliers that could dominate the result.
4. Calculate r.
5. Optionally test the null hypothesis that the population correlation is zero.
6. Interpret both magnitude and direction in the context of the subject matter.

An Original Example

Imagine an instructor records weekly practice hours and final quiz scores for 10 students. Students who practiced more tended to score higher, though not perfectly. After calculation, the result is r = 0.82 with p = 0.004.

This indicates a strong positive linear relationship in the sample. As practice hours increase, quiz scores also tend to increase. Still, the correlation does not prove that practice alone caused the higher scores. Other variables, such as prior background or motivation, may also contribute.

How to Interpret the Size of r

There is no universal cutoff that fits every field, but many researchers use rough guidelines:

• around 0.10: weak linear relationship
• around 0.30: moderate linear relationship
• around 0.50 or higher: strong linear relationship

These are only starting points. In some research areas, a small correlation can still be important if the outcome is costly or difficult to change.

Common Mistakes

• Using Pearson correlation when the pattern is curved rather than linear.
• Ignoring a single outlier that changes the conclusion.
• Interpreting correlation as proof of cause and effect.
• Using Pearson correlation for ordinal or heavily skewed data when another method would be better.

Short Summary

Pearson correlation is one of the clearest ways to describe a linear relationship between two continuous variables. A simple scatterplot, careful assumption checking, and context-aware interpretation will help you use it correctly.