How to Report a Regression in APA 7 (With a Table)

Regression is the most used multivariate analysis in psychology and also one of the most poorly reported. Here's what must appear in APA 7, how to build the regression table, and the mistakes a reviewer spots instantly. The general guide is in how to report results in APA 7.

What a regression report must include

In a multiple regression in APA 7 you report two levels. First, the overall model: the R² (and adjusted R²) and its significance through the F test with its degrees of freedom. Second, each predictor: the unstandardized coefficient B with its standard error, the standardized coefficient β, the t-value and the p-value. Predictors usually go in a table, with the model summarised in the text.

If you already have those coefficients in your output (SPSS, R, Jamovi) and what's missing is the formatting, the APA 7 regression table generator builds the full table for you, with B, SE, β, t, p and the 95% CI, ready to paste into the manuscript.

Example in the text

"The model explained a significant proportion of variance in well-being, R² = .34, adjusted R² = .32, F(3, 196) = 33.70, p < .001. Self-esteem was the strongest predictor, B = 0.45, SE = 0.08, β = .39, t(196) = 5.63, p < .001, followed by social support, B = 0.22, SE = 0.07, β = .21, t(196) = 3.14, p = .002."

Remember the formatting: R, F, B, SE, t and p in italics; β (beta, a Greek letter) upright. R² keeps the leading zero; bounded values like β or p don't (β = .39, not 0.39). And if what you actually have are the three tables SPSS produces and you cannot see where each number comes from, drop them into the output interpreter and it breaks them down in this same format.

The regression table

For more than two predictors, the table is essential for clarity. The standard columns are: Predictor, B, SE B, β, t, p and, increasingly required, the 95% confidence interval of B (95% CI [lower, upper]). APA format: horizontal lines at the top, below the headers and at the bottom, no vertical lines, and a footnote with the model's R².

Multicollinearity, hierarchical models and common mistakes

If you check multicollinearity, mention the variance inflation factors (VIF) in the text or a note. In a hierarchical regression, report the change in R² (ΔR²) and its F test at each step, because that's what justifies the approach. The most penalised mistakes: giving only the β values without the Bs and their standard errors, omitting the model's R² or F test, interpreting a coefficient without "holding the other predictors constant", and reporting p as the only information while ignoring effect size.

Frequently asked questions

How do you report a multiple regression in APA 7?

On two levels. First the model: R² and adjusted R² with the F test and its two degrees of freedom, R² = .34, adjusted R² = .32, F(3, 196) = 33.70, p < .001. Then each predictor with B, its standard error, the 95% CI of B, β, t and the exact p. With three or more predictors the coefficients belong in a table and the model summary stays in the text.

How do you report a hierarchical regression in APA 7?

Step by step. For each block you give the R² of the model up to that point and, above all, the change: ΔR² with its F test, its two degrees of freedom and its p. For example: "Adding social support in Step 2 significantly increased the variance explained, ΔR² = .08, ΔF(1, 196) = 23.76, p < .001." That increment is the whole reason for running the model in steps, so it cannot be left out; the usual layout is one column per step with the coefficients of the final model underneath.

What columns does an APA 7 regression table have?

Predictor, B, SE B, the 95% CI of B, β, t and p. Horizontal rules at the top, under the headers and at the bottom, no vertical rules, and a note underneath with N, R², adjusted R² and the model's F with its degrees of freedom. In a hierarchical model, add the ΔR² of each step to that note.

Do I report B or β in a regression?

Both. B is expressed in the original units of the variables, which is what lets you interpret the coefficient substantively and put a confidence interval on it; β is standardized and lets you compare the relative weight of predictors within the same model, but it is not comparable across samples with different variances. Reporting only the β values, with no Bs and no standard errors, is one of the most repeated reviewer comments on a regression.

Which regression values take a leading zero and which do not?

R² and p cannot exceed 1, so they lose it: R² = .34, p = .002. B, SE, t and F can exceed 1 and keep it: B = 0.45, t(196) = 5.63. β is a Greek letter, so it is not italicised, and it is conventionally written without the zero (β = .39) because it sits on a correlation-like scale.

Do I have to report multicollinearity (VIF)?

It is not a compulsory statistic, but a line in the text or in the table note is increasingly expected: "All variance inflation factors were below 2.5." The classic VIF > 10 threshold and the stricter VIF > 5 are conventions, not tests, so what counts is that you checked and that you say what you found. If two predictors are strongly correlated, discuss it when you interpret their coefficients instead of treating the VIF as a pass/fail check.

That's your regression reported without flags. If the model gets complex, interactions, mediation, hierarchical data, we go through it together in my statistical consulting and I leave it ready to paste.