How to Read a Correlation Matrix

A correlation matrix shows every pair of variables' relationship at once. Here's how to read the grid — and how Stratum highlights the relationships that actually matter.

When you have several numeric variables, a correlation matrix shows the correlation between every pair in one grid. It's the fastest way to see the structure in wide data before you build a model.

What you're looking at

Each cell is the correlation (−1 to +1) between the variable in its row and the one in its column. The diagonal is always 1 (every variable correlates perfectly with itself), and the grid is symmetric, so you only need to read one half. Stratum can highlight the cells that are statistically significant (by p-value), so the relationships worth your attention stand out from the noise.

A correlation matrix in Stratum with significant correlations highlighted — Every pair's correlation, with the significant ones highlighted.

How to read it fast

Start with the highlighted cells — those are the statistically significant relationships worth investigating.
Watch the target row — if you're predicting one variable, its row shows which others relate to it.
Spot redundancy — two predictors that correlate strongly with each other are a multicollinearity warning for any regression you build.

Numbers hide shapes. Correlation only captures linear (Pearson) or monotonic (Spearman) association. A strong cell still deserves a scatter plot to confirm the relationship is real and roughly linear — not driven by a single outlier.

In Stratum

The Correlation analysis builds the matrix with Pearson or Spearman coefficients, p-values, confidence intervals, and significance highlighting so the relationships that matter are easy to find. To see every pair as a picture, pair it with a scatter-plot matrix.

Download Stratum

Frequently asked questions

What does a correlation matrix show?

The correlation between every pair of numeric variables in a dataset, arranged in a grid. The diagonal is always 1, and the matrix is symmetric, so each pair appears once on each side.

How do I read a correlation matrix?

Read each cell as the correlation (−1 to +1) between its row and column variable. Focus on the cells Stratum highlights as statistically significant, and check your target variable's row.

How does a correlation matrix help with multicollinearity?

Predictors that correlate strongly with each other confuse a regression. The matrix flags those redundant pairs before you model — confirm with VIF afterward.

Pearson or Spearman for a correlation matrix?

Pearson for clean linear relationships; Spearman when the data is skewed or relationships are monotonic but curved. Stratum offers both.