Section 3.2

Reporting Statistics in APA Style

Q: My output has no confidence interval on the effect size. Do I leave it out?

Check again first, because SPSS is inconsistent about it. Since version 27, its one-way ANOVA has printed an ANOVA Effect Sizes table with a 95% interval on η², and the independent-samples t-test prints one on Cohen's d. But GLM Univariate, the route you take for a factorial design, gives partial η² with no interval at all. JASP prints one on η² and ω² from the ANOVA's effect-size options, at 90% rather than 95%, so the two packages can give different-looking intervals for the same result. If neither gives you one, the interval can be computed from the statistic and its degrees of freedom alone, and the APA Results Formatter does this. Do not invent one. An effect size bounded at zero has no symmetric ± interval, so ±2 standard errors around η² is not a valid interval.

Q: What is the difference between N and n?

Capital N is the whole sample; lowercase n is a piece of it. So a study of 120 people split into two conditions reports N = 120 once, then n = 60 per group when it describes them separately. Both are Latin, so both are italic, and both keep the distinction in every table heading and figure note as well. The place it matters most is chi-square, which carries the total inside its parentheses: χ²(1, N = 120) = 6.30. A lowercase n there would tell a reader you had tested only one cell. Similarly, whenever a subgroup analysis drops people, give its own n so the reader does not assume the full sample was used.

Q: My department uses a different style guide. Do these rules still apply?

APA sets the formatting. What you must report is the same under any style guide. Which letters go italic, whether a number keeps its leading zero, and where the degrees of freedom go are conventions that other style guides set differently, so follow whichever one your department names. Every guide agrees on what has to be there: the test statistic, its degrees of freedom, an exact p-value rather than a threshold, an effect size, and an interval around the estimate. A reader cannot judge your result without those five, whatever the formatting. If your department publishes no guide at all, use APA and say so once in your methods.

APA 7 sets exactly how a t-test, an ANOVA, a correlation, or a regression coefficient is written: which letters are italic, whether a number gets a leading zero, how a p-value is rounded. Once you know a handful of patterns, reporting statistics becomes a matter of filling in your own numbers, and you stop losing marks on format. This lesson gives those patterns, plus four rules that prevent most of the common mistakes.

The master patterns

Every result follows a template: the statistic, its degrees of freedom in parentheses, the value, then the p-value and an effect size. Learn these five and you can report almost any analysis you'll run in a thesis:

Analysis	APA 7 form
t-test	t(58) = 3.72, p < .001, d = 0.96
One-way ANOVA	F(2, 87) = 5.42, p = .006, η² = .11
Chi-square	χ²(1, N = 120) = 6.30, p = .012, V = .23
Correlation	r(58) = .34, p = .008
Regression	b = 0.42, SE = 0.14, β = .29, t(96) = 3.00, p = .003; model R² = .18, F(3, 96) = 7.02, p < .001

Italics: Latin yes, Greek no

This is the rule students break most often. A letter that stands for a statistic gets italicized when it's Latin, and stays upright when it's Greek:

Italic (Latin): t, F, r, p, M, SD, d, N, n, R², b, SE, t, z, U, V.
Upright (Greek): α, β, χ², η², ρ, ω. That makes it R² (Latin, italic) but η² (Greek, upright); β for a standardized coefficient stays upright even though the b beside it is italic.

Subscripts and the operators don't get italicized either. The "df" label, the parentheses, the "=" sign, and units all stay in normal type.

Leading zeros and decimals

Drop the zero before the decimal point only for numbers that can't exceed 1 in absolute value (p-values, correlations, and proportions). Everything that can be bigger than 1 keeps its leading zero:

p = .03, r = .34, β = .29: no leading zero (bounded by 1).
t = 3.72, M = 28.4, d = 0.96: keep the zero (d can, and often does, exceed 1).

Report most statistics to two decimals and p-values to two or three. Extra decimals add nothing: t = 3.7241 tells the reader no more than t = 3.72.

The p-value rules

Three habits keep p-values accurate and in APA form:

Report the exact value when you can: p = .023, not "p < .05". The exact number is more informative and lets a reader judge for themselves.
Floor tiny values at .001. Anything smaller is written p < .001, because you never claim more precision than that.
Never write p = .000. Software prints that when the value is too small to display, but a probability is never exactly zero. Translate it to p < .001.

Put the SD beside the mean. When you mention a group's average, give its spread right beside it, in parentheses: "the nap group (M = 28.4, SD = 4.1)". Report the descriptives before the test that compares them, so the reader sees the data before the test result.

Where the confidence interval goes

APA 7 also asks for an interval around the estimate wherever one can be computed. The p-value says whether you can rule out zero. The interval says which values are still plausible, and that is usually what your reader wants to know. From Output to Results Section shows where the interval goes in the sentence. This section shows its format.

The form is 95% CI [lower, upper]. Square brackets, the two limits separated by a comma, the confidence level in front. When every interval in your paper is a 95% one, say so the first time and then stop repeating it: 95% CIs [−7.2, 4.3], [9.2, 12.4], and [−1.2, −0.5]. Inside a table the convention differs, and Tables & Figures covers it: APA gives the interval its own LL and UL columns with the level in the header, so the brackets disappear.

Two things trip people up. The first is that the interval inherits the leading-zero rule of whatever it bounds, so the three patterns from the table above come out differently:

d = 0.96, 95% CI [0.42, 1.49]. Keeps its zeros, because d is unbounded.
r(58) = .34, 95% CI [.09, .55]. Drops them, because a correlation cannot pass 1.
b = 0.42, SE = 0.14, 95% CI [0.14, 0.70]. Keeps them: an unstandardized slope is in the outcome's own units and can be any size at all.

The second is naming what the interval is on. An interval around a mean difference is measured in points, seconds, or millimeters; an interval around d is measured in standard deviations; an interval around r is on the correlation scale. Writing "95% CI [1.59, 7.01]" without saying it is the difference in word pairs leaves the reader guessing at the units, so APA writes the quantity in: 95% CI of the difference [1.59, 7.01].

In student work, not every effect size gets an interval. η², R² and Cramér's V have intervals, but computing them needs a noncentral distribution, not the ordinary formula. So ANOVA and chi-square results are often reported without one. That is a habit, not a rule. JASP gives you a confidence interval on η² and ω² from the ANOVA's effect-size options. The Correlation & Regression Calculator prints one beside every r it computes, and the APA Results Formatter builds one for all six patterns in the table above. Means, mean differences, correlations, and regression coefficients should always have one.

The three variance-explained measures have one surprising feature. They take a 90% interval, not a 95% one, and JASP prints 90% by default. Their test is one-tailed, because only a large F or χ² counts as evidence against the null. A 90% two-sided interval is the one that matches the 5% one-sided test you ran: its lower limit moves above zero exactly when p drops below .05. A 95% interval stays at zero until p drops below .025. So a result with p between .025 and .05 is significant but gets a 95% interval starting at exactly zero, which looks like a contradiction but is not one. For the ANOVA in the table, the two come out as 90% CI [.02, .21] and 95% CI [.01, .23]. Report the 90% interval, and it agrees with the significance test.

Hunt the errors

Below is a results paragraph a student handed in. It contains eleven APA violations: wrong italics, stray leading zeros, a "p = .000", a missing N, and more. Click any part you think breaks a rule. Correct clicks fix the text and reveal the rule. A few correctly formatted statistics are mixed in too, so check before you click.

🔍 Find the 11 APA errors

Click the statistics that look wrong. Green = fixed; the rule appears below. Dotted underlines mark the clickable parts.

APA style is a checklist. Every rule here is mechanical: italicize the Latin letters, leave the Greek upright, drop leading zeros only on bounded numbers, floor tiny p-values, and put the SD next to its mean. Run that checklist over your Results section once and you'll catch nearly every formatting error a grader would mark.

Keep that checklist on your desk: print the one-page APA reporting cheat sheet, or convert your own numbers with the APA formatter.

Why it matters: correct notation lets a reader parse your numbers at a glance. Italic t says "test statistic"; the df in parentheses tells them your sample size and design; the missing leading zero on p says "this can't be bigger than 1". Correct typography also shows that you know what each number is. When you need a d, r, or η² to report, the effect-size converter uses the same conventions.

Problem 100 of the practice problems gives you a results sentence whose arithmetic is correct but which breaks three APA 7 rules.

Common questions

My output has no confidence interval on the effect size. Do I leave it out?

Check again first, because SPSS is inconsistent about it. Since version 27, its one-way ANOVA has printed an ANOVA Effect Sizes table with a 95% interval on η², and the independent-samples t-test prints one on Cohen's d. But GLM Univariate, the route you take for a factorial design, gives partial η² with no interval at all. JASP prints one on η² and ω² from the ANOVA's effect-size options, at 90% rather than 95%, so the two packages can give different-looking intervals for the same result. If neither gives you one, the interval can be computed from the statistic and its degrees of freedom alone, and the APA Results Formatter does this. Do not invent one. An effect size bounded at zero has no symmetric ± interval, so ±2 standard errors around η² is not a valid interval.

What is the difference between N and n?

Capital N is the whole sample; lowercase n is a piece of it. So a study of 120 people split into two conditions reports N = 120 once, then n = 60 per group when it describes them separately. Both are Latin, so both are italic, and both keep the distinction in every table heading and figure note as well. The place it matters most is chi-square, which carries the total inside its parentheses: χ²(1, N = 120) = 6.30. A lowercase n there would tell a reader you had tested only one cell. Similarly, whenever a subgroup analysis drops people, give its own n so the reader does not assume the full sample was used.

My department uses a different style guide. Do these rules still apply?

APA sets the formatting. What you must report is the same under any style guide. Which letters go italic, whether a number keeps its leading zero, and where the degrees of freedom go are conventions that other style guides set differently, so follow whichever one your department names. Every guide agrees on what has to be there: the test statistic, its degrees of freedom, an exact p-value rather than a threshold, an effect size, and an interval around the estimate. A reader cannot judge your result without those five, whatever the formatting. If your department publishes no guide at all, use APA and say so once in your methods.