How to Report a Logistic Regression in APA 7 (With Example)

A binary logistic regression predicts a two-category outcome (pass or fail, yes or no) from one or more predictors. The write-up trips people up because the coefficients are on the log-odds scale, so you have to report odds ratios and their confidence intervals, not just the raw B. This guide gives you the exact APA 7 format, a worked example you can copy, and the mistakes reviewers catch.

What you need before you write a single word

An APA 7 write-up of a binary logistic regression needs these values from your output:

1. An overall model test, usually the likelihood-ratio chi-square with its degrees of freedom and p value.
2. A pseudo R-squared for the model, such as Nagelkerke or McFadden.
3. For each predictor: the coefficient B, its standard error (SE), the Wald statistic (or z), and the p value.
4. The odds ratio for each predictor, reported as Exp(B) or OR, with its 95% confidence interval.
5. Optionally, the model's classification accuracy (percent correctly classified).

The APA 7 format template

Report the model, then the individual predictors:

A binary logistic regression was run to predict [outcome] from [predictors]. The model was statistically significant, chi-square(df) = X.XX, p = .XXX, Nagelkerke R-squared = .XX. [Predictor] was a significant predictor, B = X.XX, SE = X.XX, Wald = X.XX, p = .XXX, OR = X.XX, 95% CI [X.XX, X.XX].

Formatting rules reviewers actually check:

• Italicize the symbols: B, SE, OR, p, R.
• Report the odds ratio with a 95% confidence interval, because the interval shows precision and whether the effect could be null (an interval crossing 1).
• Keep the leading zero on B, SE, and OR (they can exceed 1). Drop the leading zero on p and on the pseudo R-squared (which cannot exceed 1).
• Round B, SE, and OR to two decimals; report p to two or three decimals, and if it is below .001, write p < .001.
• Interpret the odds ratio, not the raw B: an OR of 1.50 means the odds of the outcome are 1.5 times higher for each one-unit increase in the predictor.

A worked example

Suppose you predicted whether students passed a course (1 = pass, 0 = fail) from weekly study hours and attendance rate, with 200 students.

• Model: chi-square(2) = 42.15, p < .001, Nagelkerke R-squared = .28; 78% correctly classified.
• Study hours: B = 0.41, SE = 0.09, Wald = 20.76, p < .001, OR = 1.51, 95% CI [1.26, 1.80].
• Attendance: B = 0.03, SE = 0.02, Wald = 2.25, p = .134, OR = 1.03, 95% CI [0.99, 1.07].

Written up in APA 7, that becomes:

A binary logistic regression predicted course outcome (pass vs fail) from weekly study hours and attendance rate. The model was statistically significant, chi-square(2) = 42.15, p < .001, Nagelkerke R-squared = .28, and correctly classified 78% of cases. Study hours significantly predicted passing, B = 0.41, SE = 0.09, Wald = 20.76, p < .001, OR = 1.51, 95% CI [1.26, 1.80]: each additional weekly study hour was associated with 1.51 times higher odds of passing. Attendance was not a significant predictor, B = 0.03, SE = 0.02, p = .134, OR = 1.03, 95% CI [0.99, 1.07].

The APA 7 table (recommended for logistic regression)

A predictor table is the clearest way to present the coefficients:

Note. Outcome coded 1 = pass, 0 = fail. Model chi-square(2) = 42.15, p < .001, Nagelkerke R-squared = .28.

Mistakes reviewers catch

• Reporting only B and no odds ratio. The odds ratio is what readers interpret. Always report Exp(B) with its confidence interval.
• Omitting the confidence interval. APA 7 expects the 95% CI around the odds ratio.
• Treating the pseudo R-squared like an OLS R-squared. Nagelkerke and McFadden are not proportion-of-variance-explained; name which one you used and interpret it loosely.
• Interpreting the raw B directly. B is a change in log-odds, which few readers think in. Convert to an odds ratio.
• A leading zero on p or the pseudo R-squared, or dropping the leading zero on OR (which can exceed 1).
• Forgetting the overall model test. Report the model chi-square before the individual predictors.

Before you report: did the test's assumptions hold?

Binary logistic regression assumes:

• A binary outcome and independent observations.
• Linearity of the logit for continuous predictors (the relationship between each continuous predictor and the log-odds is linear).
• No severe multicollinearity among predictors (check the VIF).
• An adequate sample size, often stated as at least 10 outcome events per predictor.

For predicting a continuous outcome instead, use multiple regression. For the association between two categorical variables with no prediction, use the chi-square test of independence. See the assumptions guide for how to check each condition.

Let KyroStat do the write-up for you

Odds ratios, Wald statistics, confidence intervals, and a pseudo R-squared are easy to transpose wrong by hand. KyroStat fits the logistic regression on your data, checks the assumptions, and hands you the finished APA 7 sentences, a predictor table with odds ratios and confidence intervals, the plots, and the underlying Python or R code. Upload your spreadsheet, and the report is done in seconds.

Frequently asked questions

Do I report B or the odds ratio? Report both, but interpret the odds ratio. B is the log-odds coefficient; Exp(B) is the odds ratio readers actually understand.

Which pseudo R-squared should I use? Nagelkerke is the most commonly reported because it scales to a maximum of 1. McFadden is also acceptable. Name whichever you report and do not interpret it as variance explained.

What does an odds ratio below 1 mean? The predictor lowers the odds of the outcome. An OR of 0.70 means the odds are 30% lower for each one-unit increase.

How big a sample do I need? A common rule of thumb is at least 10 events (cases in the rarer outcome category) per predictor, though more is better for stable estimates.