How to Interpret Regression Output in SPSS (Worked Example)

Q: What does R squared mean in regression?

R squared is the proportion of the variation in your outcome that your predictors explain together. For example, R² = .28 means the model explains 28% of the variation in the outcome. What counts as a good value depends heavily on your field.

Q: What is the difference between R squared and adjusted R squared?

Adjusted R squared corrects R squared for the number of predictors in the model, so it does not automatically rise when you add more variables. It is especially useful when comparing models or when the sample is small.

Q: What does the ANOVA table in regression tell me?

It tests whether your model as a whole predicts the outcome better than using the mean alone. A significant F test (a small p-value) suggests that at least one predictor is related to the outcome.

Q: What is the difference between B and Beta?

B is the unstandardised coefficient: the predicted change in the outcome for a one-unit increase in the predictor, in the original units. Beta is the standardised coefficient, which puts predictors on a common scale so you can compare their relative importance.

Q: What does a p-value of less than .05 mean for a coefficient?

It suggests that, if the true coefficient were zero, a result this large would be unlikely. Most dissertations use .05 as the threshold, but always report the exact p-value and the effect size rather than relying on significance alone.

Q: What should I do if a predictor is not significant?

Report it honestly. A non-significant predictor is still a finding. Do not remove variables just to produce a cleaner model, and discuss possible reasons, such as sample size, in your discussion chapter.

Q: How do I check multicollinearity?

Look at the collinearity statistics, such as the Variance Inflation Factor (VIF), in SPSS. Common rules of thumb treat values above about 5 to 10 as a concern, but check what your supervisor or field expects.

Q: Does a significant regression prove cause and effect?

No. Regression shows association and prediction. Causal claims need an appropriate design, such as an experiment, and careful justification.

SPSS gives you several tables when you run a regression, and it is not obvious which numbers matter. This guide shows you which ones to read, what they mean and how to write them up. It covers linear regression, the most common type in dissertations.

The 3 Tables That Matter

A linear regression in SPSS produces three main tables. Read them in this order:

1. Model Summary

Shows how much of the variation in your outcome the model explains (R and R²).

2. ANOVA

Tests whether the model as a whole predicts the outcome.

3. Coefficients

Shows the effect of each predictor: its size (B), its significance (Sig.) and its confidence interval.

📌 Need your regression run and written up? Get a data analysis quote →

1. Model Summary: R and R²

R squared (R²) is the share of the variation in your outcome explained by all your predictors together. R² = .28 means 28%.
Adjusted R² corrects for the number of predictors. Use it, or report both, especially with small samples or several predictors.
What counts as "good" depends on your field. Behavioural and social science models often have lower R² than physical science models, so compare with similar studies.

2. The ANOVA Table

This tests whether your model, taken as a whole, predicts the outcome better than simply using the average. Look at the F value and its Sig. (p-value). If p is below your threshold (usually .05), the model is statistically significant. Report it as F(df regression, df residual) = value, p = value.

3. The Coefficients Table

This is the table you will discuss most. For each predictor, look at:

Column	What it tells you
B (unstandardised)	The predicted change in the outcome for a one-unit increase in the predictor, holding the other predictors constant. It is in the original units.
Std. Error	How precise the estimate of B is. Smaller is more precise.
Beta (standardised)	A unit-free version of B, useful for comparing the relative strength of predictors measured on different scales.
t and Sig.	A test of whether that predictor's coefficient differs from zero. A small p-value suggests a relationship.
95% CI for B	The range of plausible values for the coefficient. If it does not include zero, that agrees with a significant p-value.
(Constant)	The predicted outcome when all predictors equal zero. It is often not meaningful on its own.

Worked Example

This example is fictional, and the numbers are invented to show the structure. Suppose you predict a final exam score (out of 100) from weekly study hours and attendance (%), using 100 students.

Model summary: R² = .28, adjusted R² = .27.
ANOVA: F(2, 97) = 18.86, p < .001.

Predictor	B	SE	Beta	t	p	95% CI for B
(Constant)	38.40	—	—	—	—	—
Study hours (per week)	1.33	0.36	.40	3.69	< .001	[0.62, 2.04]
Attendance (%)	0.21	0.08	.25	2.62	.010	[0.05, 0.37]

How to read it:

The model explains about 28% of the variation in exam scores and is statistically significant.
Holding attendance constant, each extra weekly hour of study is associated with a predicted 1.33-point higher exam score.
Holding study hours constant, each extra percentage point of attendance is associated with a predicted 0.21-point higher score.
Study hours has the larger Beta (.40 vs .25), so it appears to be the stronger predictor in this model.

Assumptions to Check

Linear regression relies on assumptions, so check and report them briefly. In SPSS you can check most through the plots and statistics options:

Linearity: the relationship between each predictor and the outcome is roughly a straight line
Independence of errors: observations are independent of one another
Homoscedasticity: the spread of the residuals is roughly even across predicted values
Normality of residuals: the residuals are roughly normally distributed
No serious multicollinearity: predictors are not too highly correlated with each other (check VIF)
Outliers and influential cases: a few extreme values are not driving the result

Not sure whether regression is the right test in the first place? See our guide on which statistical test to use for your dissertation.

How to Report It

A short APA-style write-up of the example above might read:

A multiple linear regression was conducted to predict exam score from weekly study hours and attendance. The model was significant, F(2, 97) = 18.86, p < .001, and explained 28% of the variance (R² = .28, adjusted R² = .27). Study hours (B = 1.33, SE = 0.36, β = .40, p < .001) and attendance (B = 0.21, SE = 0.08, β = .25, p = .010) were both significant predictors.

Add the table to your chapter and keep interpretation for the discussion. Our guide to writing a dissertation results chapter explains how to structure this. Once the results are written up, the discussion chapter is where you interpret them against the literature, and our complete dissertation guide shows how the chapters fit together.

Common Mistakes

⚠️ Claiming cause and effect. A significant regression shows an association and predictive relationship, not proof that one variable causes another.

Reporting only p-values without B, confidence intervals or R²
Confusing B and Beta, or comparing B values across predictors measured in different units
Ignoring the assumptions
Adding many predictors to a small sample to inflate R²
Over-interpreting the constant when zero is not a realistic value of the predictors
Pasting SPSS screenshots instead of a formatted table

Frequently Asked Questions

What does R squared mean in regression?

R squared is the proportion of the variation in your outcome that your predictors explain together. For example, R² = .28 means the model explains 28% of the variation in the outcome. What counts as a good value depends heavily on your field.

What is the difference between R squared and adjusted R squared?

Adjusted R squared corrects R squared for the number of predictors in the model, so it does not automatically rise when you add more variables. It is especially useful when comparing models or when the sample is small.

What does the ANOVA table in regression tell me?

It tests whether your model as a whole predicts the outcome better than using the mean alone. A significant F test (a small p-value) suggests that at least one predictor is related to the outcome.

What is the difference between B and Beta?

B is the unstandardised coefficient: the predicted change in the outcome for a one-unit increase in the predictor, in the original units. Beta is the standardised coefficient, which puts predictors on a common scale so you can compare their relative importance.

What does a p-value of less than .05 mean for a coefficient?

It suggests that, if the true coefficient were zero, a result this large would be unlikely. Most dissertations use .05 as the threshold, but always report the exact p-value and the effect size rather than relying on significance alone.

What should I do if a predictor is not significant?

Report it honestly. A non-significant predictor is still a finding. Do not remove variables just to produce a cleaner model, and discuss possible reasons, such as sample size, in your discussion chapter.

How do I check multicollinearity?

Look at the collinearity statistics, such as the Variance Inflation Factor (VIF), in SPSS. Common rules of thumb treat values above about 5 to 10 as a concern, but check what your supervisor or field expects.

Does a significant regression prove cause and effect?

No. Regression shows association and prediction. Causal claims need an appropriate design, such as an experiment, and careful justification.

Need Help With Your Regression Analysis?

Our postgraduate writers run the analysis in SPSS, R and more, and write it up as a clear results chapter linked to your research questions.

Get a Data Analysis Quote →

How to Interpret Regression Output in SPSS (With a Worked Example)