A point-biserial correlation measures the relationship between a continuous variable and a variable with exactly two categories, such as attended or did not attend, or treatment or control. It is an ordinary Pearson correlation in which the two-category variable is coded 0 and 1, so it is reported almost the same way. The one thing to add is which group is coded 1, because the sign depends on it.
What you need before you write a single word
- The coding of the two-group variable (which group is 1).
- The correlation rpb and its degrees of freedom (N − 2).
- The exact p value, and ideally a 95% CI.
- Each group's mean and SD, which make the result readable.
The APA 7 format template
A point-biserial correlation showed a significant [positive] relationship between [group variable] (coded 0 = [group A], 1 = [group B]) and [outcome], rpb(df) = .XX, p = .XXX. [Group B] (M = X.XX, SD = X.XX) scored higher than [group A] (M = X.XX, SD = X.XX).
All 53 templates, ready to paste with the italics set: the Rescue Vault, $27 →
Formatting rules reviewers actually check:
- State the coding: a positive r means the group coded 1 scored higher.
- No leading zero on r or p; df in parentheses.
- Add the group means; a correlation alone hides how big the difference is in the outcome's units.
A worked example
Eighty students sat a statistics exam; 40 had attended an optional revision session (coded 1) and 40 had not (coded 0).
- Attended: M = 66.95, SD = 11.74
- Did not attend: M = 60.27, SD = 11.06
- rpb(78) = .28, p = .011
Written up in APA 7:
A point-biserial correlation showed a significant positive relationship between attending the revision session (0 = did not attend, 1 = attended) and exam score, rpb(78) = .28, p = .011. Students who attended (M = 66.95, SD = 11.74) scored higher than students who did not (M = 60.27, SD = 11.06).
The same data in an independent t-test give t(78) = 2.62, p = .011: an identical p, because the two tests are mathematically equivalent.
Mistakes reviewers catch
- No coding stated. Without it, the sign of r is meaningless.
- Using it for a variable that was cut into two groups. If the two groups came from splitting a continuous score (for example, high vs low anxiety), the biserial correlation, or better the original continuous variable, is appropriate.
- Reporting it and a t-test as separate findings. They are the same result; choose the one your field expects.
Let KyroStat do the write-up for you
KyroStat recognises two-group variables, runs the correlation or the equivalent t-test, and writes the sentence with the coding and the group means.
Frequently asked questions
Point-biserial or independent t-test? They give the same p. Use the independent t-test when the question is about group differences, and the point-biserial when it sits in a table of correlations.
Does it go in a correlation table? Yes. Note the coding in the table note. See how to make an APA correlation table.
What about two continuous variables? Use Pearson's correlation.