SPSS command: SPSS REGRESSION

Linear regression in your browser

Linear regression models a scale outcome as a weighted sum of predictors and estimates how much the outcome changes with each one, holding the others fixed.

Try with example data Run on your own file

Free · No account · Files stay on your device · Updated 18 October 2026

What it tells you

Crosstabs runs SPSS REGRESSION on your device with Enter, Stepwise, Forward, Backward and Remove blocks, R² change between blocks, standardized coefficients, VIF and the usual residual diagnostics.

How much variance in the outcome the predictors explain (R², adjusted R²), and how much each added block contributes (R² change with its F test).
Whether the model as a whole beats the mean-only model (ANOVA F test).
The expected change in the outcome per unit of each predictor (B with its 95% confidence interval) and which predictors matter most on a common scale (Beta).
Whether predictors overlap too much (tolerance, VIF, condition indices) and whether any case fits badly (casewise diagnostics, Cook's distance, leverage).

When to use it

Use it when

Explaining or predicting a continuous outcome (spend, satisfaction score, time) from several predictors.
Testing whether a set of variables adds explanatory power beyond controls, by entering them in a second block.
Key-driver analysis in market research, using Beta and part correlations to compare drivers.

Use something else when

A yes/no outcome: use binary logistic regression; for three or more categories, multinomial or ordinal regression.
Counts or strongly skewed positive outcomes: generalized linear models with a Poisson, negative binomial or gamma distribution fit better.
Interaction terms or factorial designs: the panel enters main effects only, so build product variables first or use GLM Univariate.
Survey data from a stratified or clustered design: use Complex Samples GLM for valid standard errors.

How to run it in Crosstabs

01
Choose Linear regression
In Statistics mode open Regression → Linear regression (REGRESSION) and pick the Dependent (scale) variable.
02
Fill block 1
In Block 1 choose a Method (Enter, Stepwise, Forward or Backward) and add predictors with Add independent variable.
03
Code categorical predictors
For each predictor set Coding to Scale covariate or Dummy-code (categorical) and pick the Reference category. Dummies are created explicitly; nothing is dummy-coded automatically.
04
Add more blocks if needed
Click Add block for hierarchical models; later blocks can also use Remove to drop variables entered earlier. Adjust PIN (.05) and POUT (.10) under Stepping criteria.
05
Declare the weight and run
If the workspace has a weight, choose Frequency weight, Sampling weight or Run unweighted, then click Run regression.

Run Linear Regression in your browser

Worked example

Computed by the Crosstabs engine from the synthetic data below when this page was built — not typed in.

Try with example data

Data

60 synthetic stores with monthly sales (thousands), floor area (m²), advertising spend (thousands) and a region flag (urban or suburban). Sales are generated as a linear function of all three plus normal noise; the model enters floor area first, then adds advertising and region in a second block.

Model Summary

Model	R	R Square	Adjusted R Square	Std. Error of the Estimate	R Square Change	F Change	Sig. F Change
1	0.747	0.558	0.550	15.13	0.558	73.15	<.001
2	0.905	0.819	0.810	9.84	0.261	40.50	<.001

Durbin-Watson (final model) = 1.808. Dependent Variable: sales.

ANOVA (Model 2)

Item	Sum of Squares	df	Mean Square	F	Sig.
Regression	24598.4	3	8199.5	84.60	<.001
Residual	5427.4	56	96.9
Total	30025.7	59

Coefficients (Model 2)

Item	B	Std. Error	Beta	t	Sig.	95% CI for B	VIF
(Constant)	39.154	5.281	—	7.41	<.001	28.57 to 49.73	—
area	0.183	0.014	0.740	12.90	<.001	0.15 to 0.21	1.02
ads	3.569	0.539	0.380	6.62	<.001	2.49 to 4.65	1.02
region = urban	13.616	2.601	0.302	5.24	<.001	8.41 to 18.83	1.03

The region dummy is coded 1 = urban, 0 = suburban (the chosen reference).

How to read it

Floor area alone explains 55.8% of the variance in sales. Adding advertising and region raises R² by 0.261 (F change 40.50, p < .001), and the full model is significant, F(3, 56) = 84.60, p < .001. Holding area and region constant, each extra thousand spent on advertising is associated with 3.57 thousand more in monthly sales (95% CI 2.49 to 4.65). The largest VIF is 1.03, so collinearity is not a concern.

Output tables explained

Model Summary: R, R², adjusted R², standard error of the estimate and, per model, the change statistics (R² change, F change, df1, df2, Sig. F change); Durbin-Watson for the final model.
ANOVA: Regression and residual sums of squares, df, mean squares, F and its significance for each model.
Coefficients: B and its standard error, Beta, t, Sig., the 95% confidence interval for B, zero-order, partial and part correlations, tolerance and VIF.
Excluded Variables: For stepwise methods and multi-block models: Beta In, t, Sig., partial correlation and tolerance of the variables not in each model.
Collinearity Diagnostics: Eigenvalues, condition indices and variance proportions per model. A condition index above 30 with two or more large proportions signals harmful collinearity.
Residuals Statistics: Minimum, maximum, mean and SD of predicted values, residuals, studentized and deleted residuals, Mahalanobis and Cook's distance and centered leverage; Casewise Diagnostics lists cases with |standardized residual| > 3.

Assumptions and notes

The outcome is linearly related to each predictor, and residuals have constant variance; check the residual statistics and casewise list.
Residuals are independent (Durbin-Watson near 2 for data in time order) and roughly normal for small samples.
Predictors are not near-duplicates: a VIF above about 10 or a tolerance below .1 makes coefficients unstable. Variables below tolerance .0001 are not entered.
Missing values are removed listwise. Frequency weights must be whole numbers; a sampling weight gives point estimates only and disables Stepwise, Forward and Backward.

How the results are validated

The REGRESSION engine is tested against independent reference values (Python 3.13.13, NumPy 2.5.4, SciPy 1.18.1, statsmodels 0.15.0, pandas 2.3.3). With Enter, B, SE, t, p, CI, Beta, zero-order, partial and part correlations, tolerance, VIF, R², adjusted R², F, the ANOVA sums of squares and Durbin-Watson match statsmodels OLS within 1e-9 (relative to the larger of 1 and the reference value); per-case influence measures match OLSInfluence within 1e-9, residual statistics within 1e-8 and Belsley collinearity diagnostics within 1e-8 (variance proportions 1e-7). B, SE, p, VIF and adjusted R² also agree with R 4.6.1 lm and car 3.1.5 vif within 1e-9. Further tests cover frequency weights as case replication, R² change and F change for nested blocks and Remove, and Stepwise, Forward and Backward sequences replayed by a hand-run F-to-enter procedure in the generator. Validation evidence

Questions

How do I include a categorical predictor such as region?

Add it and set Coding to Dummy-code (categorical), then choose the reference category. Each other category becomes a 0/1 dummy compared with the reference; in stepwise methods each dummy is treated as a separate variable, as in SPSS.

What is the difference between B and Beta?

B is in the outcome's units per unit of the predictor, which is what you use for prediction. Beta is the same effect in standard-deviation units, which lets you compare predictors measured on different scales.

Should I use stepwise selection?

Use it for exploration, not confirmation: the p values after selection are optimistic. For hypothesis tests, enter predictors chosen in advance, in blocks if order matters.

Can I save predicted values or residuals to my data?

Not from this panel. It reports Residuals Statistics and Casewise Diagnostics for the final model, with the row number of each flagged case, but does not write new variables to the dataset.

What happens with a sampling weight?

You get weighted coefficients and R², but standard errors, tests and intervals are withheld because ordinary least-squares inference is invalid under a survey design. Use Complex Samples GLM for design-based inference.

Run Linear Regression on your own data

Open a .sav, CSV or Excel file. Everything runs in your browser; the file never leaves your device.

Open the workspace

SPSS is a trademark of IBM Corp. Crosstabs is not affiliated with or endorsed by IBM.