How to Interpret Multiple Regression Output in SPSS (and Report It in APA 7)

A multiple regression asks two questions: does the set of predictors explain the outcome, and what does each predictor add while the others are held constant? SPSS answers the first in the Model Summary and ANOVA tables and the second in the Coefficients table, where unstandardized and standardized coefficients sit side by side and are easy to confuse. This guide reads each table cell by cell with a four-predictor worked example, lists the misreadings to avoid, and gives the APA 7 write-up. The reporting format alone is in the multiple regression APA guide.

What SPSS gives you

Analyze > Regression > Linear, with all predictors in Independent(s), Method: Enter, and Confidence intervals and Collinearity diagnostics ticked under Statistics, produces:

1. Variables Entered/Removed: the predictors and the method.
2. Model Summary: variance explained.
3. ANOVA: the test of the whole model.
4. Coefficients: one row per predictor.
5. Collinearity Diagnostics: eigenvalues and condition indices (rarely needed; the VIF column in Coefficients is usually enough).

Footnote letters (ᵃ, ᵇ) point to "Dependent Variable: Job satisfaction" and to "Predictors: (Constant)" followed by the four predictor names.

The worked example

A survey of 120 employees measured job satisfaction (0 to 100) and four predictors: autonomy (1 to 7, M = 4.62, SD = 1.08), workload (hours per week, M = 43.90, SD = 7.20), salary (in thousands of dollars, M = 61.09, SD = 13.46) and tenure (years in the job, M = 5.31, SD = 3.78). Job satisfaction averaged 50.60 (SD = 11.44). All tables below come from one dataset.

Table 1: Model Summary

• R = .627: the multiple correlation between the observed satisfaction scores and those the model predicts.
• R Square = .393: the four predictors together explain 39.3% of the variance in job satisfaction.
• Adjusted R Square = .371: R² corrected for the number of predictors. R² can only go up when you add a predictor, even a useless one; adjusted R² penalises that. Report both in a multiple regression.
• Std. Error of the Estimate = 9.066: the typical prediction error, in satisfaction points.

Table 2: ANOVA

The test of the whole model: do the predictors, together, explain more than chance?

• Regression df = number of predictors = 4. Residual df = N − predictors − 1 = 115.
• F = 18.584, Sig. = .000 (p < .001): the model is significant.
• A significant model does not mean every predictor is significant. That is the next table.

Table 3: Coefficients

SPSS groups B and Std. Error under "Unstandardized Coefficients", Beta under "Standardized Coefficients", the CI under "95.0% Confidence Interval for B" and Tolerance and VIF under "Collinearity Statistics".

Each predictor row is that predictor's effect with the other three held constant.

• B (unstandardized): the change in satisfaction for a one-unit change in the predictor. One more point of autonomy goes with 5.10 more satisfaction points; one more weekly hour of work with 0.41 fewer; one thousand dollars more salary with 0.14 more; tenure makes essentially no difference (−0.006).
• Std. Error: the uncertainty of B. t = B / Std. Error.
• Beta (standardized): the same effects in SD units, the change in satisfaction (in SDs) for a one-SD change in the predictor. Because every predictor is on the same scale here, Beta is the column for comparing them: autonomy (.484) matters most, then workload (−.258), then salary (.168).
• t and Sig.: whether each B differs from zero. Autonomy, workload and salary are significant; tenure is not (p = .979). The df for these t tests is the Residual df, 115.
• 95% CI for B: the plausible range of each effect. Tenure's interval (−0.45 to 0.44) spans zero comfortably.
• Tolerance and VIF: multicollinearity checks. VIF = 1 / Tolerance. All VIFs here are about 1.05 to 1.10, so the predictors overlap very little. VIFs above 5 (some use 10), or tolerances below .2 (or .1), mean predictors overlap so much that their individual B values are unstable.

B and Beta tell different stories when the scales differ. Salary's B (0.142) looks tiny next to autonomy's (5.100), but salary is measured in thousands of dollars with an SD of 13.46, and autonomy on a 1 to 7 scale with an SD of 1.08. Per SD, salary's effect (.168) is about a third of autonomy's (.484), not a thirty-sixth. Use B to say what one unit is worth, and Beta to compare predictors.

Sig. = .001 is not always p = .001. Workload's Sig. reads .001, but its exact p is .0008, which SPSS rounds up. Double-click the table in the Viewer and then the cell to see more decimals; report p < .001 here.

Misreadings that cost marks

• Comparing predictors by B. B depends on each predictor's units. Compare by Beta.
• Reading a significant model as "all predictors matter". The ANOVA tests the set; each predictor has its own Sig. in Coefficients.
• Reporting R Square alone. With several predictors, add adjusted R².
• Interpreting the constant. It is the predicted satisfaction for someone with 0 autonomy, 0 workload, 0 salary and 0 tenure: an impossible person. Use it only in the equation.
• Forgetting "holding the others constant". Each B is a partial effect. Salary's effect here is what salary adds over autonomy, workload and tenure.
• Writing p = .000 or trusting a rounded .001. Write p < .001, and check cells near .001 for their exact value.
• Skipping VIF. High multicollinearity can make a real predictor look non-significant.
• Dropping non-significant predictors and re-running without saying so. Report the model you planned; if you simplified it, report both and why.

The APA 7 format template

A multiple regression showed that [predictors] together explained a significant proportion of variance in [outcome], F(df1, df2) = X.XX, p = .XXX, R² = .XX, adjusted R² = .XX. [Predictor] was a significant positive/negative predictor, b = X.XX, 95% CI [X.XX, X.XX], β = .XX, p = .XXX.

The write-up for this example

A multiple regression showed that autonomy, workload, salary and tenure together explained a significant proportion of the variance in job satisfaction, F(4, 115) = 18.58, p < .001, R² = .39, adjusted R² = .37. Holding the other predictors constant, satisfaction was higher with greater autonomy, b = 5.10, 95% CI [3.50, 6.70], β = .48, p < .001, and higher salary, b = 0.14 per $1,000, 95% CI [0.02, 0.27], β = .17, p = .027, and lower with heavier workload, b = −0.41 per weekly hour, 95% CI [−0.65, −0.17], β = −.26, p < .001. Tenure did not predict satisfaction, b = −0.01, 95% CI [−0.45, 0.44], β = −.00, p = .979. Multicollinearity was low (all VIFs < 1.2).

With four or more predictors, a coefficients table (b, 95% CI, SE, β, t, p) usually replaces most of that prose; the multiple regression APA guide shows the layout.

Let KyroStat do the write-up for you

Already have the output? Paste the Model Summary and ANOVA tables from the SPSS Viewer into the box on the multiple regression reporting guide to get the APA sentence for the model. Or upload the spreadsheet to KyroStat, which runs the regression, checks residuals and multicollinearity, and writes the APA 7 sentences and the coefficients table with the code behind every number.

Also useful for the write-up (affiliate links): Proofreading for your thesis or paper (20% off with code PAP20): Paperpal

Frequently asked questions

What is the difference between B and Beta in SPSS? B (unstandardized) is the change in the outcome per one unit of the predictor, in the original units. Beta (standardized) is the change in SDs per one-SD change in the predictor, so predictors on different scales can be compared.

Which coefficient do I report? B with its 95% CI for each predictor, plus Beta (β) when readers need to compare predictors. Many APA tables show both.

The model is significant but one predictor is not. Is that a problem? No. The model test asks whether the predictors explain the outcome together; each predictor's t asks whether it adds anything over the others. Report both.

What VIF is too high? Common thresholds are VIF above 5 or above 10 (tolerance below .2 or .1). Below about 2, as here, multicollinearity is not a concern.

Why does SPSS show .001 when my p is smaller than .001? SPSS rounds to three decimals, so .0008 shows as .001. Double-click the cell to see the exact value, and report p < .001 if it is below .001.

Which related guides might I need? Often these: simple linear regression output, hierarchical regression and moderation analysis.