Pearson & Spearman Correlation Coefficient Calculator

When to use this test

• •You have two continuous (or at least ordinal) variables measured on the same cases.
• •You want to know the strength and direction of their relationship, not to predict one from the other.
• •Use Pearson r when both variables are roughly normal and you expect a straight-line relationship.
• •Use Spearman rho when variables are skewed, contain outliers or are measured on a rank scale.
• •You are screening many variables — a correlation matrix gives you a quick overview before modelling.

How to use it

1. 1

   Upload or paste your dataset

   Load a CSV or XLSX, or paste rows into the grid. You need at least two numeric columns. Pairs with missing values on either variable are dropped automatically; a missing-value summary is shown before the calculation runs.

2. 2

   Pick Pearson, Spearman or a correlation matrix

   Under the Correlation category choose Pearson (linear, normal data) or Spearman (ranked or non-normal data). To screen many variables at once, pick the correlation matrix option and select all numeric columns.

3. 3

   Assign the two variables

   Drag the two columns you want to correlate into the Variable X and Variable Y slots. Order does not affect r. For the matrix version, multi-select any number of numeric columns.

4. 4

   Read r, the p-value and the scatter plot

   DeepStats reports the coefficient, its 95% confidence interval, a p-value for the test that the true correlation is zero, and sample size. A scatter plot with an overlaid linear fit (for Pearson) makes it easy to spot non-linearity or outliers.

5. 5

   Export the table or scatter plot

   Save the correlation matrix as CSV or the scatter plot as PNG. When you run the matrix version, cells are colour-coded so strong correlations stand out at a glance.

Example with sample data

Eight data points relating study hours to exam score. Paste into the grid and run Pearson correlation to reproduce the numbers below.

Hours,Score
1,52
2,61
3,68
4,74
5,80
6,83
7,88
8,95

The calculator returns r ≈ 0.995, 95% CI [0.973, 0.999], p < 0.001, n = 8. A near-perfect positive linear relationship — as study hours rise, scores rise almost proportionally. The scatter plot shows the fit line passing cleanly through every point, confirming Pearson is appropriate.

How to interpret the results

You will see: correlation coefficient (r or rho), 95% confidence interval, two-tailed p-value, sample size (after dropping missing pairs), scatter plot with fitted line.

The coefficient (r or rho)

Ranges from −1 to +1. Zero means no monotonic relationship. Conventional benchmarks: |r| ≈ 0.1 weak, 0.3 moderate, 0.5 strong, 0.7+ very strong. Sign indicates direction — positive means variables rise together, negative means one rises as the other falls.

95% confidence interval

A range of plausible values for the true population correlation. If the interval excludes zero, the relationship is significant at the 0.05 level. Narrow intervals indicate precise estimates; wide intervals signal you need more data.

p-value

The probability of seeing a correlation at least as large as yours if the true correlation were zero. Below 0.05 is the usual threshold. Significance alone does not say the relationship is strong — in a large sample a tiny r can be significant.

Sample size

DeepStats reports n after listwise deletion of missing pairs. The precision of any correlation estimate grows with sample size; below n ≈ 20 the confidence interval gets very wide.

Scatter plot

Always inspect the scatter plot before interpreting r. A small r can hide a strong non-linear relationship, and a large r can be driven by a single outlier. Pearson is especially vulnerable to extreme values.

Assumptions

• Linearity (Pearson only).Pearson measures linear association. A U-shaped or exponential relationship will produce a low r even though the variables are strongly related. If the scatter plot shows curvature, switch to Spearman or fit a non-linear model.
• Bivariate normality (Pearson only).Both variables should be roughly normally distributed and free of extreme outliers. For skewed data or ordinal scales, Spearman rho is the appropriate alternative.
• Monotonicity (Spearman).Spearman only requires the relationship to be monotonic — as X increases, Y consistently increases or decreases. It is far less sensitive to outliers than Pearson.
• Independent pairs.Each (X, Y) pair must come from a different subject or event. Repeated measurements on the same person require a mixed-model correlation or partial correlation.
• Correlation is not causation.A strong r can reflect a common cause, reverse causation or pure coincidence. Experimental design, not correlation magnitude, establishes causality.

Related calculators

Frequently asked questions