33  Measures of Association

Measures of association describe the relationship between two variables — does rainfall move with yield? Does pest pressure move against it?

  • Correlation: measures the strength and direction of the association between two variables.
    • Pearson’s correlation: measures the strength of a linear relationship.
    • Spearman’s rank correlation: measures the strength of a monotonic relationship (not necessarily linear).
  • Covariance: measures the degree to which two variables vary together, in the original units of both variables — which makes it harder to interpret directly than correlation, since its size depends on the scale of the data.

Difference Between Covariance and Correlation

Correlation is essentially a standardized version of covariance — it rescales covariance to always fall between -1 and 1, which is what makes it comparable across datasets measured in different units. Read more on this distinction →

33.1 Example in R

Rainfall (mm) and yield (tons/ha) across 15 farms in a district:

33.1.1 Scatter Plot

A correlation above 0.9 and a scatter plot with points hugging the regression line both point to a strong, close-to-linear relationship between rainfall and yield in this dataset.


33.2 Positive Correlation

A positive correlation means two variables tend to rise and fall together. Fertilizer application and yield are a natural agricultural example, up to the point of diminishing returns.


33.3 Negative Correlation

A negative correlation means one variable tends to rise as the other falls. Pest incidence and yield are a natural example: the more a field is affected by pests, the lower its yield tends to be.

A correlation close to -1 here confirms what the downward-sloping scatter plot already shows: as pest incidence rises, yield falls, consistently and closely to linearly, across these twelve farms.


Summary

Concept Description
Measures of Association
Correlation Measures the strength and direction of association between two variables, scaled to fall between -1 and 1
Covariance Measures how two variables vary together, in the original units of the data
Positive Correlation Both variables tend to rise and fall together, e.g., fertilizer use and yield
Negative Correlation One variable tends to rise as the other falls, e.g., pest incidence and yield