Guide

From SPSS Output to APA Results

SPSS prints far more numbers than you will report. For your thesis you need to know which of the numbers on screen belong in your paragraph, which row to read, and how to format the sentence. This guide covers the five most common analyses. For each one it gives the menu path, a mock of the output with the reportable numbers highlighted, the usual misreading, and the APA 7 sentence. Three of the five use the practice datasets, so you can reproduce those numbers yourself.

Three rules before any output

Sig. means p, and it is never .000. SPSS rounds to three decimals, so .000 means p < .0005. Write p < .001, never p = .000. (Double-clicking a value in the output viewer reveals its full precision.)
Typography follows fixed rules. Latin statistical symbols are italic (t, F, p, r, d, M, SD, N); Greek ones are not (χ², η², β as a population parameter, though sample standardized β is commonly italicized). Statistics that cannot exceed 1 (p, r, η², Cramér's V) get no leading zero: p = .047, not 0.047.
A result has four parts: the descriptives, the test statistic with its df, the p-value, and an effect size. A sentence with only a p is incomplete, and Effect Size & Power explains why. The full rules are in Reporting Statistics in APA Style. The APA results formatter builds any of the sentences below from your own numbers and warns you if your p does not match your statistic.

1 · Independent-samples t-test

Menu: Analyze → Compare Means → Independent-Samples T Test. Data: sleep-experiment.csv from the practice datasets, 40 sleep-restricted vs 40 normally rested people, counting attention lapses (the same example our JASP guide runs, so you can compare the two programs side by side).

Group Statistics

condition	N	Mean	Std. Deviation	Std. Error Mean
restricted_sleep	40	22.77	6.02	.952
normal_sleep	40	20.07	5.92	.937

Independent Samples Test

	Levene's F	Sig.	t	df	Sig. (2-tailed)	Mean Diff.	Std. Error Diff.	95% CI Lower	95% CI Upper
Equal variances assumed	.095	.758	2.022	78	.047	2.700	1.335	.042	5.358
Equal variances not assumed			2.022	77.980	.047	2.700	1.335

Read Levene's columns first, and only to choose your row: Levene's p = .758 is above .05, so the top row (equal variances assumed) applies. Had Levene's been significant, you would read the bottom (Welch) row. Many statisticians use the Welch row every time.

The trap: that first Sig. column belongs to Levene's test, not your t-test. Reporting it as your p-value is one of the most common SPSS mistakes. Your p is under Sig. (2-tailed). The lesson explains what each piece means. SPSS also gives Cohen's d if you tick Estimate effect sizes (here d = 0.45).

Sleep-restricted participants made more attention lapses (M = 22.77, SD = 6.02) than rested participants (M = 20.07, SD = 5.92), t(78) = 2.02, p = .047, d = 0.45, 95% CI of the difference [0.04, 5.36].

2 · One-way ANOVA

Menu: Analyze → Compare Means → One-Way ANOVA; under Options tick Homogeneity of variance test and Welch. Data: study-methods.csv, 105 students, three revision techniques, one exam score.

Test of Homogeneity of Variances

	Levene Statistic	df1	df2	Sig.
exam_score	3.615	2	102	.030

Levene's p = .030 < .05, so the equal-variances assumption fails, and the plain ANOVA row below is not the one to report.

ANOVA

exam_score	Sum of Squares	df	Mean Square	F	Sig.
Between Groups	2200.30	2	1100.15	11.566	.000
Within Groups	9701.83	102	95.12
Total	11902.13	104

Robust Tests of Equality of Means

	Statistic	df1	df2	Sig.
Welch	13.104	2	66.43	.000

Because Levene's test was significant, report the Welch row (note its fractional df₂). Sig. .000 becomes p < .001.

Effect size: recent SPSS versions offer η² in the options. From the table it is SS_between / SS_total = 2200.30 / 11902.13 = .185. Follow-ups: a significant omnibus F tells you the means are not all equal, but not which ones differ. The post-hoc table shows which pairs differ. With unequal variances, use Games–Howell rather than Tukey. Post-Hoc Tests explains the multiple-comparisons problem these tests solve, and the one-way ANOVA lesson covers the full logic.

Exam scores differed across revision techniques, Welch's F(2, 66.43) = 13.10, p < .001, η² = .19. Because Levene's test indicated unequal variances (p = .030), Welch's robust test is reported.

3 · Pearson correlation

Menu: Analyze → Correlate → Bivariate. Data: screen-time.csv, 60 people's daily screen hours and sleep quality.

Correlations

		screen_hours	sleep_quality
screen_hours	Pearson Correlation	1	−.205
	Sig. (2-tailed)		.115
	N	60	60

The matrix is symmetric, so each pair appears twice, and the diagonal is all 1s (each variable with itself). You need three numbers: r, its p, and N.

Two traps. First, the df in your sentence is N − 2 = 58, not N. SPSS does not print it, and examiners check it. Second, and more serious: this r is misleading. One influential participant (11.5 hours of screen time, excellent sleep) pulls the correlation toward zero. Without that point, r goes from −.21 to −.40. So plot every correlation before you report it (Graphs → Chart Builder, or paste both columns into the correlation calculator, which also computes the leave-one-out range). Then decide what to do about the point and report both versions. A non-significant correlation can be reported as it is, with its confidence interval, and not as "a trend."

Screen time and sleep quality were not significantly correlated, r(58) = −.21, p = .115. One influential observation attenuated the association; excluding it, r(57) = −.40, p = .002.

4 · Chi-square test of independence

Menu: Analyze → Descriptive Statistics → Crosstabs; under Statistics tick Chi-square and Phi and Cramér's V; under Cells tick Expected. Example: 120 students. Did those who attended a revision workshop (38 of 60) pass a methods exam more often than those who didn't (24 of 60)?

Chi-Square Tests

	Value	df	Asymptotic Sig. (2-sided)
Pearson Chi-Square	6.541	1	.011
Continuity Correction	5.640	1	.018
Likelihood Ratio	6.602	1	.010
N of Valid Cases	120

Report the Pearson Chi-Square row, and read the footnote SPSS prints under the table. If any cell's expected count is below 5, the chi-square approximation may be inaccurate. Here the smallest expected count is 29, so there is no problem.

The trap: the table has three plausible-looking rows. Pearson is the standard one to report. The continuity correction is an older, conservative adjustment for 2×2 tables that most current advice leaves out. The likelihood ratio is a different statistic. Report Pearson, say in plain words what the cells showed, and add Cramér's V from the Symmetric Measures table as the effect size. Chi-Square Tests explains the full logic.

Workshop attendees passed more often (63%) than non-attendees (40%), χ²(1, N = 120) = 6.54, p = .011, Cramér's V = .23.

5 · Simple linear regression

Menu: Analyze → Regression → Linear. Example: 85 first-year students; weekly exercise hours predicting wellbeing (1–7 scale). SPSS prints regression output as three tables, and your sentence needs numbers from each:

Model Summary

Model	R	R Square	Adjusted R Square	Std. Error of the Estimate
1	.436	.190	.180	.988

ANOVA

Model	Sum of Squares	df	Mean Square	F	Sig.
Regression	19.000	1	19.000	19.469	.000
Residual	81.000	83	.976
Total	100.000	84

Coefficients

Model	B	Std. Error	Beta	t	Sig.
(Constant)	1.860	.421		4.418	.000
exercise_hours	.353	.080	.436	4.412	.000

Model Summary → how much variance the model explains (R²). ANOVA → whether the model as a whole predicts better than the mean alone (F, with df 1 and 83). Coefficients → the slope's size (B, in the original units), its standardized version (Beta), and its own test (t). The (Constant) row is the intercept. You rarely interpret it, and you almost never need its p-value.

The trap: both .000s here become p < .001. Students often copy p = .000 from this table into a thesis. Also, do not confuse B (unstandardized: +0.35 wellbeing points per extra hour) with Beta (in SD units, comparable across predictors). Simple Linear Regression explains the difference, and Multiple Regression covers models with more than one predictor. Before you trust any of it, check the residual plots.

Exercise significantly predicted wellbeing, b = 0.35, SE = 0.08, β = .44, t(83) = 4.41, p < .001; the model explained 19% of the variance, R² = .19, F(1, 83) = 19.47, p < .001.

The classic misreadings, in one list

Reporting Levene's Sig. as your p. Levene's columns only tell you which row (t-test) or which table (ANOVA → Robust Tests) to report.
Copying Sig. = .000. Write p < .001.
Wrong df. A correlation's df is N − 2; a t-test's df must match your group sizes. Examiners recompute these.
A p-value with no effect size. Add d, η², r itself, Cramér's V, or R². The effect-size converter converts between them.
Trusting a coefficient you never plotted. One influential point can hide (or invent) a relationship.
Interpreting in the Results section. The sentence states what was found. What it means belongs in the Discussion (From Output to Results Section).

Check your numbers: paste any statistic and its df into the APA results formatter. It rebuilds the sentence with correct typography and recomputes the p-value from the statistic, as statcheck does, so you catch a typo before your examiner does.

Keep going

Reporting Statistics in APA Style: the full formatting ruleset, with an error-hunt game.
From Output to Results Section: build results sentences from any output.
Analyze Your Thesis Data in JASP: the same steps in free software.
Practice Datasets: reproduce this guide's t-test, ANOVA, and correlation yourself.
Which test should I use? if you're not yet sure which of the five you need.
The Final Checklist: check that every number matches before you submit.
One Study, Start to Finish: a whole study, from the research question to the results paragraph.