Reporting Multiple Regression in SPSS
Coefficients, Model Fit and Wording
Reporting multiple regression in SPSS draws on three tables: Model Summary, ANOVA and Coefficients. Taking N after exclusions, B with its interval and R squared from the right one keeps the APA 7 table and the sentence consistent with the output.
IN SHORT
N, Fit and One Row Per Predictor
Regression results are reported in two layers: the model as a whole, with the sample size after exclusions, R squared and adjusted R squared and the F test, and then each predictor, with its coefficient B, interval, standard error and exact p-value, read with the other predictors held constant in the model. For reporting multiple regression in SPSS, take these from the Model Summary, ANOVA and Coefficients tables (Analyze, Regression, Linear; menu labels vary by release). Researchers who send their saved tables to OnlineSPSS, a specialist statistical consultancy for academic, clinical and professional research, have the table and the paragraph prepared or checked by an in-house PhD-level statistician, and the substantive reading of the coefficients remains the researcher’s.
In This Guide
ANALYSIS SUPPORT
Analysis Support for This Method
A coefficient table presupposes a finished model. Where predictors are still being selected, or are meant to enter in planned blocks whose contribution is compared, the modeling is done before the reporting.
Multiple Linear Regression
Several predictors of one continuous outcome, analyzed with the assumptions examined and the estimates reported with their intervals.
Hierarchical Regression
Predictors entered in planned blocks, so that what each block adds to the model can be tested and reported in order.
BEFORE YOU REPORT IT
What the Coefficient Table Depends On
Four decisions determine which coefficients are quoted and how they are described, and each is fixed before the table is built.
The Model Reported Output can hold several models when predictors were entered in steps. The report states which model is presented, and the table follows that model alone.
B or Beta B keeps the units of each predictor, whereas the Beta column puts predictors on a common scale. Beta is added only where the predictors are comparable and a comparison is intended.
Coding of Predictors A binary predictor coded 0 and 1 has a B equal to the difference between its two categories with the other predictors held constant, and a predictor with several categories enters as dummy variables against a named reference group.
Cases Used A case with a missing value on any variable in the model is dropped, so N after exclusions is smaller than the dataset, and the text gives both numbers.
IN SPSS
Which of the Three Tables Supplies Each Regression Value
Collinearity and residuals are examined first, in Multiple Regression in SPSS: Checks and Output, and the values below then come from the three tables that IBM SPSS prints for the model.
Take N After Exclusions
The Model Summary prints no sample size, so take N from the ANOVA table as the Total df plus 1, or from Descriptive Statistics if requested.
Check: Compare this N with the dataset size. The gap is the number of cases lost to a missing value on any variable in the model, and the text should say how many.
Take R Square and Adjusted R Square
From the Model Summary table, take R Square and Adjusted R Square, and R if the multiple correlation is wanted.
Tip: Report adjusted R square beside R square, and lean on it when comparing models of different sizes or when the sample is small relative to the number of predictors.
Take the F Test and Its df
From the ANOVA table, take F, the Regression and Residual degrees of freedom, and Sig. for the model as a whole.
Watch: The overall F tests all coefficients together. A significant F does not show that every predictor contributes, and a non-significant one is still reported in the same detail.
Take B, SE and the Interval for B
From the Coefficients table, take B, Std. Error and the 95.0% Confidence Interval for B, which is requested under Statistics in the dialog.
Check: If the interval columns are missing, rerun the model with confidence intervals ticked. Without them the table gives a point estimate and no statement of its precision.
Decide Whether Beta Belongs
Take the Standardized Coefficients Beta only where the predictors are on comparable scales and comparing them is part of the question.
Watch: Beta depends on the spread of each predictor in this sample, so it does not rank importance and does not carry over to a sample with a different spread.
Take t and p, Then Assemble
Take t and Sig. for each predictor, then build the table and the sentence from the same rows of the Coefficients table.
Check: The df for each t is the Residual df from the ANOVA table. Confirm that the table note and the sentence quote the same df and the same N.
READING THE OUTPUT
From Model Summary, ANOVA and Coefficients to the Report
The sample size, the fit statistics and the predictor rows are scattered over three tables, so each row below names the table to open and what to confirm there.
| Output value | Where SPSS prints it | Where it goes and what to check |
|---|---|---|
| Total df plus 1 | ANOVA | N after exclusions, in the sentence and the table note, beside the dataset size |
| R Square and Adjusted R Square | Model Summary | Model fit, in the sentence and the table note, with the adjusted value alongside |
| F, Regression and Residual df, Sig. | ANOVA | F(df1, df2) and p for the model as a whole, kept apart from the coefficient p-values |
| B | Coefficients, Unstandardized Coefficients | The estimate in outcome units per unit of the predictor, with its sign described in words |
| Std. Error and 95.0% Confidence Interval for B | Coefficients | The SE B column and the interval beside B, at the level stated in the note |
| Beta | Coefficients, Standardized Coefficients | An optional column, kept where predictors are comparable and a comparison is intended |
| t and Sig. | Coefficients | The p column, with df from the Residual row; a printed .000 becomes p < .001 |
| (Constant) | Coefficients | Shown in the table, interpreted only if zero is meaningful for every predictor |
REPORTING
A Coefficient Table and a Sentence Drawn From the Same Rows
The table carries the predictors and its note carries the model statistics, so the sentence and the table are built together and checked together.
A multiple regression APA table gives each predictor a row with B, its standard error, the interval for B and the p-value, and the note carries the model statistics. Reporting regression coefficients in APA style also means naming the outcome, the predictors and the sample after exclusions in the sentence, and saying that each coefficient describes an association with the other predictors in the model taken into account.
A multiple linear regression examined whether [Predictor 1], [Predictor 2] and [Predictor 3] were associated with [Outcome] among ___ participants after exclusions. The model accounted for ___% of the variance in [Outcome], R² = ___, adjusted R² = ___, F(___, ___) = ___, p = ___. With the other predictors held constant, [Predictor 1] was [positively / negatively] associated with [Outcome], B = ___, 95% CI [___, ___], t(___) = ___, p = ___.
| Table 1. Multiple Regression Predicting [Outcome] (N = ___) | |||
|---|---|---|---|
| Predictor | B [95% CI] | SE B | p |
| (Constant) | ___ [___, ___] | ___ | ___ |
| [Predictor 1] | ___ [___, ___] | ___ | ___ |
| [Predictor 2] | ___ [___, ___] | ___ | ___ |
| [Predictor 3] | ___ [___, ___] | ___ | ___ |
| Note. R² = ___, adjusted R² = ___, F(___, ___) = ___, p = ___. B = unstandardized coefficient; SE B = its standard error. Add a β column only where the predictors are comparable. | |||
B with its interval is the effect estimate for each predictor, in the units of the outcome per unit of the predictor, and it is the figure a reader can act on. R squared is the effect size for the model as a whole, and adjusted R squared corrects it for the number of predictors. For one predictor’s unique contribution, the squared Part correlation (requested under Statistics as Part and partial correlations) gives the share of variance that only that predictor accounts for. A coefficient is read against the scales of its predictor and the outcome and not against fixed labels.
COMMON MISTAKES
Regression Table Errors a Reviewer Will Find
The move from three printed tables to one APA table is where these appear, so the finished table is set against the output once more.
Beta Used to Rank Reading the largest Beta as the most important predictor treats a sample-specific scaling as a ranking. Importance depends on the question, and effects in real units are reported through B.
Model Fit Without N R squared quoted with no sample size hides how many cases the model used. N after exclusions belongs in the sentence and in the note to the table.
p-Values Without Intervals A column of p-values shows which predictors cleared a threshold but not how large the effects are or how precisely they are estimated, whereas B with its interval shows both.
Causal Verbs for Coefficients Wording such as drives or raises the outcome claims more than an observational regression shows. Coefficients describe associations, with the other predictors in the model taken into account.
Reference Group Unnamed The B of a dummy-coded predictor is a difference from a reference category. If the table does not name that category, the coefficient cannot be interpreted.
Constant Read Blindly The intercept is the predicted outcome when every predictor is zero, a point that may lie outside the data. It is shown in the table and discussed only where zero is meaningful.
Regression table built from three outputs? Independent Statistical Review compares each row with the Coefficients table.
FREQUENTLY ASKED QUESTIONS
Fit Statistics, Dummies and p-Values in a Regression Report
Which R squared to quote, how dummy variables are laid out and which of two p-values goes where must be settled before the table is finished.
Do I report R squared or adjusted R squared?
Both, with the adjusted value as the steadier guide when the sample is small relative to the number of predictors. R squared never falls when a predictor is added, so it flatters larger models. The sentence gives the two values together, and the note repeats them with the F test, so that a reader can compare models fitted to the same data.
How should dummy-coded predictors appear in the table?
One row per dummy variable, labeled with the category it stands for, and the reference group named in the note. Each B is then a difference from that reference category with the other predictors taken into account, and the sentence describes it as such, with the number of cases in each category given in the text or in a descriptive table.
Which p-value goes in the table, the t test or the F test?
The p-value of the t test goes beside each coefficient, and the p-value of the F test belongs to the model as a whole and goes in the note and the sentence. They answer different questions, and quoting one for the other misdescribes the result: a significant model can contain predictors whose own intervals include zero, and the reverse can occur.
Which numbers in a regression table most often need correcting?
Usually the N after exclusions, the df beside each t and an unnamed dummy reference category. An in-house PhD-level statistician at OnlineSPSS sets each row beside the printed tables, or builds the table from the output, within Results Reporting & Interpretation. The claims drawn from the coefficients are your own.
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
Can You Trace Each
Coefficient to a Printed Row?
A regression table is assembled from three printed tables, and a slip in one carries into the others. A table checked row by row against all three is one that a reviewer can follow from output to page.
No Commitment Required
Strictly Confidential
Fixed Project Fee
In-House PhD-Level Statisticians
Same-Day Response
Serving Researchers Since 2014