Overview
Correlation analysis is the most commonly misinterpreted statistical output in business analytics. A correlation of 0.7 between marketing spend and revenue does not mean marketing spend causes revenue. It does not mean that increasing marketing spend by 10% will increase revenue by 7%. It means the two variables tend to move together — and the reason they move together could be causation, reverse causation, a common cause, or coincidence.
The right correlation coefficient depends on the data types: Pearson for continuous normally distributed variables, Spearman for ordinal or non-normal continuous variables, Kendall's tau for small samples or many ties. Using Pearson on ordinal data produces a coefficient that is numerically wrong. Testing multiple correlations without correction inflates the false positive rate — with 20 variables, you expect one spurious significant correlation at α=0.05 even if nothing is truly correlated.
The Correlation Analysis & Interpretation Prompt generates a complete correlation framework: coefficient selection by data type, significance testing with multiple comparison correction, correlation matrix visualization specification, and an interpretation guide that correctly states what correlation implies and what it doesn't.
What you get: - Coefficient selection by data type and distribution - Significance testing with multiple comparison correction - Correlation matrix with clustering and visualization - Spurious correlation detection - Causation limitation communication template
Built for: analysts and data scientists who need defensible, correctly interpreted correlation analyses for exploratory data analysis and feature selection.