How To Calculate Correlation Between Two Variables In Stata

Stata Correlation Calculator

Paste two equal length numeric series, choose Pearson or Spearman, and instantly see the correlation coefficient, interpretation, sample size, and a chart that helps you understand the relationship before you run the command in Stata.

Interactive Calculator

Tip: In Stata, the most common commands are correlate x y for Pearson and spearman x y for Spearman. This calculator helps you verify the relationship before you run your data in Stata.

Expert Guide: How to Calculate Correlation Between Two Variables in Stata

Correlation is one of the most widely used measures in statistics because it tells you whether two variables tend to move together and how strongly they do so. In Stata, learning how to calculate correlation between two variables is straightforward, but interpreting the result correctly requires more than just typing a command. You need to understand which correlation statistic fits your data, what assumptions apply, how missing values affect the sample, and how to interpret the sign and magnitude of the coefficient.

At its core, correlation summarizes the direction and strength of association between two variables. If the coefficient is positive, higher values of one variable tend to occur with higher values of the other. If it is negative, higher values of one variable tend to occur with lower values of the other. The closer the coefficient is to 1 or negative 1, the stronger the linear or ranked association. A value near 0 suggests weak or no consistent relationship, although that does not always mean the variables are unrelated because some non linear patterns can still produce a small correlation.

In Stata, researchers usually work with two main options. Pearson correlation is the standard choice for continuous variables with an approximately linear relationship. Spearman rank correlation is often preferred when the data are ordinal, heavily skewed, or affected by outliers, or when the relationship is monotonic rather than strictly linear. The calculator above lets you evaluate both approaches quickly before you move into your Stata workflow.

What correlation means in practical terms

Suppose you are studying the relationship between study hours and exam performance. A positive correlation would mean students who study more tend to score higher. Suppose instead you analyze price and quantity demanded. A negative correlation may appear if higher prices are associated with lower purchase volume. Correlation does not prove causation, but it is often the first quantitative step in understanding a relationship.

• Positive correlation: both variables rise together.
• Negative correlation: one variable rises while the other falls.
• Near zero correlation: little consistent linear or monotonic association.
• Strong correlation: the data points cluster closely around an upward or downward pattern.
• Weak correlation: the points are more dispersed and less predictable.

Basic Stata commands for two variable correlation

If your data are already loaded into Stata and your variables are named x and y, the standard syntax is simple.

correlate x y

This command produces the Pearson correlation coefficient. If you want a rank based alternative, use:

spearman x y

If you want significance levels and pairwise sample handling for a wider matrix of variables, you may also use:

pwcorr x y, sig

These commands seem similar, but they answer slightly different questions. correlate is often used for a standard correlation matrix. pwcorr is useful when missing data differ by variable pair. spearman transforms values into ranks, reducing sensitivity to extreme values and non normality.

Step by step: calculating correlation in Stata

1. Load or import your dataset. For example, use use, import excel, or import delimited depending on your source.
2. Inspect the variables. Run summarize and codebook to verify scale, range, and missing values.
3. Check the relationship visually. A scatter plot is often the best first step. In Stata, use twoway scatter y x.
4. Choose the right correlation method. Use Pearson for continuous data with a roughly linear pattern, or Spearman if the relationship is monotonic or if outliers are a concern.
5. Run the command. Use correlate x y or spearman x y.
6. Interpret the coefficient. Focus on sign, magnitude, sample size, and whether the relationship makes substantive sense.
7. Report the result clearly. Include the coefficient, the sample size, the method used, and if relevant the p value or confidence interpretation.

How Pearson and Spearman differ

The most common mistake is choosing Pearson automatically without checking assumptions. Pearson measures linear association using the raw values. Spearman replaces the raw values with ranks and measures how consistently one variable increases or decreases as the other changes.

Method	Best for	Relationship type	Outlier sensitivity	Typical Stata command
Pearson correlation	Continuous variables with a roughly linear pattern	Linear association	Higher sensitivity	correlate x y
Spearman rank correlation	Ordinal or skewed variables, monotonic trends	Monotonic association	Lower sensitivity than Pearson	spearman x y

As a rule of thumb, if a scatter plot shows a straight line tendency without major outliers, Pearson is often appropriate. If the pattern is consistently increasing but curved, or if the data contain several unusually extreme values, Spearman may better represent the association.

Example with realistic statistics

Imagine a sample of 120 students. You are analyzing weekly study hours and final exam score. After cleaning the data and checking for obvious entry errors, you run Pearson correlation in Stata. You obtain a coefficient of 0.62. This suggests a moderately strong positive linear association. Students with more study hours generally tend to score higher, though the relationship is not perfect and other factors likely matter.

Now imagine a second scenario where you study neighborhood income rank and access to preventive care rank across 80 local areas. Because ranks are more appropriate than raw skewed measures, Spearman correlation produces 0.71. That indicates a strong positive monotonic relationship: areas with higher income rank also tend to have better preventive care rank.

Research example	Sample size	Method	Coefficient	Interpretation
Study hours vs exam score	120	Pearson	0.62	Moderately strong positive linear relationship
Income rank vs preventive care rank	80	Spearman	0.71	Strong positive monotonic relationship
Price vs quantity purchased	250	Pearson	-0.48	Moderate negative relationship

Interpreting correlation magnitude carefully

People often look for fixed labels such as weak, moderate, or strong, but context matters. In physics, a coefficient of 0.30 might be considered small. In social science or public health, 0.30 can be meaningful, especially if the phenomenon is influenced by many variables. Practical interpretation depends on the field, sample size, measurement quality, and the decision context.

• 0.00 to 0.19: often described as very weak
• 0.20 to 0.39: often described as weak
• 0.40 to 0.59: often described as moderate
• 0.60 to 0.79: often described as strong
• 0.80 to 1.00: often described as very strong

These categories are only rough conventions. A coefficient of 0.45 in a noisy real world dataset may be quite important. A coefficient of 0.90 can still be misleading if the relationship is driven by just a few points or by a coding problem.

Why your scatter plot matters

Never rely on the coefficient alone. A scatter plot can reveal outliers, clusters, curved relationships, and data entry errors that a single number hides. In Stata, a useful first graph is:

twoway (scatter y x) (lfit y x)

This overlays a fitted line on the scatter, helping you judge whether a linear summary is appropriate. If the plot curves upward and then levels off, Pearson may understate or misdescribe the association. If there are a few extreme points far from the main pattern, Pearson may be distorted. That is one reason Spearman can be valuable as a robustness check.

Handling missing values in Stata

Missing values are easy to overlook. Stata commands may differ in how observations are used. With correlate, the analysis typically uses observations that are complete across the specified variables. With pwcorr, pairwise deletion can be used so each variable pair is calculated using all available cases for that pair. This can produce larger usable samples but also means coefficients in a matrix are not always based on the same observations.

If you are working with exactly two variables, the practical issue is simpler: ensure both variables are nonmissing for the observations you want in the analysis. You can check this with commands like:

count if missing(x) | missing(y) summarize x y if !missing(x) & !missing(y)

When not to use simple correlation

Correlation is useful, but it is not universal. You should be cautious in several situations:

• If one variable is binary and the other continuous, other measures may be more informative depending on the research design.
• If the relationship is clearly non linear, the coefficient may underrepresent a strong association.
• If time series data are involved, autocorrelation and trends can create misleading correlations.
• If both variables are influenced by a third variable, the simple bivariate correlation may be confounded.
• If causality is your goal, regression, experimental design, or causal inference methods are usually needed.

Useful Stata workflow for better analysis

A strong workflow combines descriptive statistics, visualization, and formal analysis. Here is a practical sequence many analysts use:

summarize x y graph matrix x y twoway (scatter y x) (lfit y x) correlate x y spearman x y

This sequence gives you summary ranges, a visual check, the standard Pearson coefficient, and a rank based alternative. If Pearson and Spearman are very different, inspect the data more closely because outliers or non linearity may be driving the difference.

How to report the result in a paper or report

When reporting a correlation from Stata, include the method, coefficient, and sample size. If inferential reporting is required, also include the significance level. A concise example is:

Pearson correlation showed a positive association between study hours and exam score, r = 0.62, n = 120.

For ranked data:

Spearman rank correlation indicated a strong positive association between income rank and preventive care rank, rho = 0.71, n = 80.

If your audience is not statistically specialized, also explain the practical meaning in plain language. For example, state that students who studied more tended to earn higher scores on average.

Common mistakes to avoid

1. Using Pearson without checking whether the scatter plot is approximately linear.
2. Ignoring outliers that can heavily shift the coefficient.
3. Assuming correlation implies causation.
4. Failing to examine missing values and sample size.
5. Interpreting a low correlation as proof of no relationship when the pattern may be curved.
6. Using aggregated data and overlooking ecological bias.

How the calculator above relates to Stata

The calculator on this page uses the same underlying ideas as Stata. For Pearson correlation, it computes covariance divided by the product of the variables’ standard deviations. For Spearman correlation, it first converts the values into average ranks when ties exist and then calculates Pearson correlation on those ranks. The scatter chart helps you see whether the relationship is roughly linear, which is exactly the kind of diagnostic step you should perform before interpreting a Stata output table.

If your calculator result shows a coefficient such as 0.853, you can expect Stata to report the same value, subject to rounding and any differences in how you treated missing values or ties. That makes the tool useful for learning, checking hand entered datasets, and validating whether your Stata syntax is returning what you expect.

Authoritative references for further reading

For deeper statistical background and software oriented guidance, consult these authoritative sources:

• UCLA Statistical Methods and Data Analytics: Stata resources
• Penn State Eberly College of Science: statistics lessons on correlation and association
• NIST Engineering Statistics Handbook

Final takeaway

To calculate correlation between two variables in Stata, start by understanding your variables, graphing the relationship, and choosing the right method. Use correlate for Pearson correlation when the relationship is approximately linear and the variables are continuous. Use spearman when ranks or monotonic trends are more appropriate. Always interpret the coefficient in context, inspect the scatter plot, and remember that association is not the same as causation. If you follow that workflow consistently, your correlation analysis in Stata will be both technically sound and substantively meaningful.