METHOD GUIDE
Pearson Correlation in SPSS
Linearity, Outliers and Reporting r With Its Interval
A Pearson r describes a straight-line association and can be bent by a single point, so the scatterplot comes before the coefficient. Linearity, influential cases and the interval decide whether the value you report means what it seems to.
IN SHORT
What r Measures and What It Leaves Unsaid
Pearson’s r measures the strength and direction of the linear association between two continuous variables, and its test asks whether that association differs from zero in the population. It describes association, not causation or agreement between two measurements. To run Pearson correlation in SPSS, open Analyze, Correlate, Bivariate, move the variables in, keep Pearson ticked, and look at a scatterplot of each pair before reading the Correlations table. OnlineSPSS, a specialist statistical consultancy for academic, clinical and professional research, carries out and reviews this analysis through an in-house PhD-level statistician.
ANALYSIS SUPPORT
Analysis Support for This Method
A coefficient answers a narrower question than a write-up often implies. Where prediction, or the size of a change in the outcome, is the aim, a regression model says more than r can.
Correlation Analysis (Pearson & Spearman)
The study of association between your variables carried out within your design, with the checks on linearity and outliers, the choice of coefficient and the interval agreed for your data.
Simple Linear Regression
A model that predicts one variable from another and gives a slope in the units of the data, for when the question moves from association to prediction.
BEFORE YOU RUN IT
What the Coefficient Needs From Your Data
Four conditions decide whether r can be interpreted as intended, and most of them are visible in a plot before any test is run.
Two Continuous Variables Both measures should be continuous or close to it. A single 5-point item suits a rank-based coefficient better, and a binary variable gives a point-biserial value that reads differently.
A Straight-Line Pattern The scatterplot should show a trend a line can describe. A curve or a set of clusters can give an r near zero, or a misleadingly high one, although the association is real in another form.
Influential Points One or two extreme points can create or hide a correlation. They are examined in the plot, and any decision to keep them, check them or report r with and without them is stated.
Independent Pairs Each pair should come from a different case. Repeated measurements from the same people, or participants nested in classes, make pairs related and the usual p-value too optimistic.
IN SPSS
How to Run Pearson Correlation in SPSS
Steps for IBM SPSS, limited to the choices that decide what the matrix can support. Menu labels and the options offered differ by version.
- Make the scatterplot linearity check in SPSS first, through Graphs and Chart Builder, and look at the pattern and the extreme points.
- Open Analyze, Correlate, Bivariate, and move the variables into the list.
- Keep Pearson ticked under Correlation Coefficients, and choose two-tailed unless a direction was stated in advance.
- Under Options, request means and standard deviations, and choose how missing values are excluded.
- Run it and read the N in each cell of the matrix before the coefficients.
- For an interval, request Confidence interval in the dialog where your version offers it, or Bootstrap where your license includes it.
READING THE OUTPUT
Reading the Correlations Table Cell by Cell
Each cell of an SPSS correlation matrix holds three numbers, and the N printed in the cell is the one most often overlooked.
| Output element | What it tells you | What to check |
|---|---|---|
| Descriptive Statistics (if requested) | The mean, standard deviation and N of each variable | That means and standard deviations go into the text, since the matrix does not show them |
| Pearson Correlation | The coefficient r, from -1 to 1, printed above and below the diagonal | That its size is read from the plot and from r squared, not against fixed labels without context |
| Sig. (2-tailed) | The p-value for the hypothesis that the population correlation is zero | That it is read as evidence of a nonzero association, never as the size of that association |
| N | The number of cases behind each cell | That each coefficient is reported with its own N, since pairwise exclusion lets N differ between cells |
| Significance asterisks | Marks for significance at .05 and .01 beside r | That they do not replace the report, and are not trusted across a large matrix without a decision on multiplicity |
| Whole matrix | All pairs of k variables at once, k(k-1)/2 distinct tests | That the number of pairs tested is stated, and that exploratory results are labeled as such |
| Bootstrap rows in the Correlations table (if requested) | Bias, standard error and a percentile or BCa interval for r | That the interval type, the number of samples and the seed are recorded |
REPORTING
Reporting r With Its Interval and a Scatterplot
The coefficient travels with its N, its interval and the scatterplot it came from, and each underscore is a value from your own matrix.
Write r(df) = ___, 95% CI [___, ___], p = ___, with df equal to the number of pairs minus 2 and the N stated. Where your version offers it, the Bivariate dialog’s Confidence interval option prints Fisher z limits for r; otherwise use Bootstrap where your license includes it, or compute the limits elsewhere. Describe the direction in words and interpret the size against what is usual in your field. Include the scatterplot, name any influential points and say whether r was computed with or without them. When many coefficients are reported, say how many were examined and how multiplicity was handled.
COMMON MISTAKES
Why a Reported Correlation Gets Queried
Reviewers question a correlation for what it hides more often than for what it shows, and the first three problems are visible in a scatterplot.
No Scatterplot A coefficient reported without its plot hides curves, clusters and outliers that change its meaning. The plot is made before the test is read, and it often belongs in the paper.
One Point Doing the Work A single extreme case can produce or cancel a strong r. Reporting r with and without it, and saying why the case was retained, shows that the result does not rest on one observation.
Restricted Range Sampling only a narrow band of one variable, such as high scorers, usually pulls r toward zero compared with the full population. The sampling and the range are described next to the coefficient.
Significance Read as Strength In a large sample a trivial r is significant, and in a small one a sizable r may not be. The coefficient and its interval carry the interpretation, and the p-value sits behind them.
Every Pair Tested A matrix with many variables throws up some significant coefficients by chance. The number of pairs examined is reported, and the unplanned results are labeled exploratory.
A Causal Reading A correlation in observational data does not show that one variable influences the other. Confounding and reverse direction remain open, and the discussion says so.
Matrix already produced? Independent Statistical Review checks the scatterplots, the intervals and the wording of each coefficient before you submit.
FREQUENTLY ASKED QUESTIONS
Questions About Pearson r in SPSS Output
Pairwise N, the interval for r and a coefficient close to zero are the three points most often misread in a first correlation matrix.
Why does the N differ between cells of my correlation matrix?
By default SPSS excludes missing values pairwise, so each coefficient uses every case that has values on both of its variables. Report each coefficient with its own N, or choose listwise exclusion in Options when one common sample is needed, accepting the loss of cases. With many missing values, the pattern of missingness matters more than the setting.
How do I get a confidence interval for r in SPSS?
Where your version offers it, the Bivariate Correlations dialog has a Confidence interval option that adds Fisher z limits for r; otherwise use the Bootstrap option where your license includes it, or compute the limits from r and N in another program. The Fisher z interval and the p-value assume approximate bivariate normality, so with skewed variables or influential points a bootstrap interval is the safer check. Report the interval with r, because it shows how precisely the sample pins down the population correlation.
Does an r near zero mean there is no relationship?
No. It means there is no linear association, and a strong curved or clustered pattern can also give a value near zero. The scatterplot shows which case applies. A narrow range on one variable can also pull r toward zero, so the sampling is checked before concluding that a relationship is absent.
Can OnlineSPSS check the correlations in my results section?
Yes. An in-house PhD-level statistician can review the plots, the choice of coefficient and the intervals behind each value you report, or carry out the analysis on your data from the start. When the analysis is finished, Results Reporting & Interpretation covers the write-up of the tables. The interpretation and the conclusions remain with you.
How do you protect my research data and confidentiality?
Your data and project details are handled confidentially, and project information is not published. Researcher names and institutions are not used publicly without permission. An NDA is available where required and can be arranged before sensitive materials are shared. Read more about confidentiality and data security.
What happens when I request an assessment, and how long does it take?
You share your research question, design and relevant materials, such as the dataset or the output you have, and a statistically trained Research Coordinator reviews the request and responds the same day. The analysis, deliverables, timeline and one fixed project fee are then confirmed before any work begins, so the timeline follows from the agreed scope.
Still have a question? Find more answers in our full FAQ.
NEXT STEP
Correlations Run and
Wording Still Unsettled?
Attach the matrix as SPSS printed it, the scatterplots if you have them and a line on how each variable was measured. From that, the checks on linearity and the interval for r can be scoped.
No Commitment Required
Strictly Confidential
Fixed Project Fee
In-House PhD-Level Statisticians
Same-Day Response
Serving Researchers Since 2014