How to Report a Chi-Square Test of Independence in APA 7

Academic Decoder (@academicdecoder)

Created by a PhD-qualified lecturer · Updated

The short answer

A chi-square test of independence tests whether two categorical variables are associated. In APA 7 you describe the pattern with counts and percentages, then report χ² with its degrees of freedom and the sample size, the exact p-value and an effect size (Cramér's V, or φ for a 2 × 2 table). You also check, and can mention, that the expected counts were large enough.

This guide is for undergraduate and master's students writing a dissertation results chapter. It uses one research scenario throughout: 150 students said whether they study full-time or part-time, and which support format they prefer: Online, In-person or Both. Each student is counted once. These are simulated data for illustration, the same example as the free SPSS Output Explorer.

When to use a chi-square test of independence

  • Two categorical variables, such as study mode (full-time, part-time) and preferred support format (Online, In-person, Both).
  • Counts of people or cases, not means, scores or percentages. Each person falls in exactly one cell of the table.
  • Each person counted once. If the same people were measured twice, for example before and after a workshop, use McNemar's test instead.

With only one categorical variable, such as whether students' preferences match an expected split, you need a chi-square goodness-of-fit test instead. If your outcome is a score, such as an exam mark, compared across groups, you need a t-test or a one-way ANOVA.

Assumptions and important checks

  • Independent observations. Each person appears in one cell only, and one person's answer does not affect another's. This comes from your design, not from a test.
  • Large enough expected counts. The expected count is how many people would fall in a cell if the two variables were unrelated. A common rule is that no more than 20% of cells should have an expected count below 5, and none below 1.
  • Counts, not percentages. Run the test on the number of people in each cell. Percentages help you describe the pattern, but the test needs the counts.

The expected count footnote most students skip

SPSS checks the expected counts for you, in footnote a under the Chi-Square Tests table. In the worked example it reads: "0 cells (0.0%) have expected count less than 5. The minimum expected count is 20.85." That means the check is passed, and you can say so in one short sentence of your write-up.

If the footnote reports too many small expected counts, the usual χ² p-value may not be reliable. Use Fisher's exact test instead (SPSS prints it for 2 × 2 tables; for larger tables, choose Exact under Exact Tests), or combine categories where that still answers your research question. Say which you did.

How to read the SPSS output

This is the worked example as it would appear in SPSS. The highlighted values are the ones you report.

Simulated SPSS output · layout modelled on SPSS 29

Study mode * Preferred support Crosstabulation (some columns hidden on small screens)
Study modeOnlineIn-personBoth
Full-time count30 (highlighted: a value you report)37 (highlighted: a value you report)15 (highlighted: a value you report)
Full-time %36.6% (highlighted: a value you report)45.1% (highlighted: a value you report)18.3% (highlighted: a value you report)
Part-time count20 (highlighted: a value you report)17 (highlighted: a value you report)31 (highlighted: a value you report)
Part-time %29.4% (highlighted: a value you report)25.0% (highlighted: a value you report)45.6% (highlighted: a value you report)
Chi-Square Tests
ValuedfAsymptotic Significance (2-sided)
Pearson Chi-Square13.786ᵃ (highlighted: a value you report)2 (highlighted: a value you report).001 (highlighted: a value you report)
Likelihood Ratio13.9752.001
N of Valid Cases150 (highlighted: a value you report)
Symmetric Measures
ValueApproximate Significance
Nominal by NominalCramer's V.303 (highlighted: a value you report).001
N of Valid Cases150
a. 0 cells (0.0%) have expected count less than 5. The minimum expected count is 20.85. SPSS shows χ² to three decimals (13.786); APA 7 reports it to two, 13.79. Not shown: the Linear-by-Linear Association row and the Phi row, which SPSS also prints.
  • The crosstab shows the pattern. Read the percentages within each row: 45.6% of part-time students preferred Both, against 18.3% of full-time students, while full-time students more often preferred In-person (45.1% against 25.0%).
  • The Pearson Chi-Square row: χ²(2, N = 150) = 13.79, p = .001. The df is (rows − 1) × (columns − 1) = (2 − 1) × (3 − 1) = 2. It shows the two variables are associated, but not which categories drive the association.
  • Cramér's V = .30, from Symmetric Measures, shows how strong the association is.

Finding which categories drive the association

A significant χ² tells you the two variables are associated somewhere in the table. To see where, compare the observed and expected counts, or ask SPSS for adjusted standardised residuals (Cells, then Adjusted standardized). As a rough guide, a residual beyond ±1.96 marks a cell with noticeably more or fewer people than expected. When you check many cells, a stricter cut-off is sometimes used, so check what your department expects.

Both

Full-time
−3.6
Part-time
+3.6
What it shows
Part-time students chose Both far more often than expected

In-person

Full-time
+2.6
Part-time
−2.6
What it shows
Full-time students chose In-person more often than expected

Online

Full-time
+0.9
Part-time
−0.9
What it shows
Close to what was expected in both groups

Key values to report, and where to find them

You need the pattern (counts and percentages), four values from the test itself, an effect size and the expected count check. Some of them need an extra option ticked in your software.

Counts and row percentages

What it tells the reader
The pattern in the table
SPSS
Crosstabulation (Cells: Observed and Row percentages)
jamovi
Contingency Tables, Cells: Observed counts and Row %
R
table()

χ²

What it tells the reader
How far the observed counts are from the expected counts
SPSS
Chi-Square Tests, Pearson Chi-Square row, Value
jamovi
χ² (Tests)
R
chisq.test(…, correct = FALSE), X-squared

df

What it tells the reader
(rows − 1) × (columns − 1)
SPSS
Same row, df
jamovi
df
R
df

N

What it tells the reader
How many people were counted
SPSS
N of Valid Cases
jamovi
N
R
sum(table)

p

What it tells the reader
Whether the association is statistically significant
SPSS
Same row, Asymptotic Significance (2-sided)
jamovi
p
R
p-value

Cramér's V or φ

What it tells the reader
How strong the association is
SPSS
Symmetric Measures (Statistics: Phi and Cramer's V)
jamovi
Statistics: Phi and Cramer's V
R
effectsize::cramers_v()

Expected count check

What it tells the reader
Whether the test can be trusted
SPSS
Footnote a under Chi-Square Tests (Cells: Expected for the full table)
jamovi
Cells: Expected counts
R
chisq.test(…)$expected

Which effect size to report

The p-value tells you whether there is evidence of an association, not how strong it is. For that you report:

  • φ (phi) for a 2 × 2 table. For example, in a 2 × 2 table of study mode by preferred learning format with 60 students, χ²(1, N = 60) = 5.45, p = .020, φ = .30.
  • Cramér's V for larger tables, such as the 2 × 3 worked example: V = .30.
  • Benchmarks depend on the table size. When the smaller side of the table has two categories, .10, .30 and .50 are the usual guides for small, medium and large. With more categories on both sides, the benchmarks are lower. In the worked example, V = .30 is a medium association.
  • Both are written without a leading zero, because they cannot exceed 1: V = .30, not 0.30.

Worked example · Chi-square test of independence

What does the chi-square test tell us?

Study mode and preferred support format were compared using simulated data from 150 students.

Example APA 7 resultAbbreviated teaching example
A chi-square test of independence showed a statistically significant association between study mode and preferred support format, χ²(2, N = 150) = 13.79, p = .001, Cramér's V = .30.

A short example for learning, not a complete, dissertation-ready results paragraph.

This tells us that study mode and preferred support format are associated, and that the association is of medium strength. It does not tell us which preferences differ between full-time and part-time students, or why. To answer the first, we need the row percentages and the adjusted residuals. The second is a question for the discussion, not the test: these data cannot show that study mode causes a preference.

What else needs reporting?

A complete write-up may need the counts and percentages for each group, a description of where the association lies, the expected count check, effect-size interpretation and appropriate qualification of the findings.

See how the Result Decoder supports the full reporting process.

Your statistical results, explained and ready to write up.

Already completed your analysis in SPSS, jamovi, JASP, R or another supported statistical program? Enter the key values from your output into Academic Decoder to understand, visualise, report and defend your findings.

  • APA 7 write-upA structured results paragraph using the values you enter.
  • Plain-English explanationUnderstand what your statistical result means, including what the effect size tells you.
  • Graphs & visualsA relevant chart of your result, downloadable as PNG or SVG where supported.
  • Discussion sentence startersDevelop your discussion in your own words.
  • Defend This ResultPractise examiner-style questions and understand the limits of your conclusions.
  • Save as Draft or FinalOrganise your results and return to them while you write your dissertation.
  • My Chapter 4Turn your saved results into an editable Word draft, with a reporting checklist.
  • Viva preparationDownload examiner-style practice questions and suggested answers to revise from.

The Decoder doesn't run the statistical analysis. You complete your analysis in your own software, then enter the relevant values from your output.

Get help writing up my chi-square results

£15.99 one-time · All 26 supported analyses · Unlimited decodes · No subscription

Reporting a non-significant result

A non-significant result is reported with the same values. Simulated data: full-time students Online 30, In-person 30, Both 22; part-time students Online 24, In-person 22, Both 22; 150 students in total.

Example APA 7 resultAbbreviated teaching example
A chi-square test of independence showed no statistically significant association between study mode and preferred support format, χ²(2, N = 150) = 0.60, p = .742, Cramér's V = .06.

A short example for learning, not a complete, dissertation-ready results paragraph.

  • Say the data did not provide sufficient evidence of an association. Do not say the variables are "independent" or "unrelated"; a non-significant test does not prove that.
  • Keep the leading zero on χ² when it is below 1 (χ² = 0.60), because χ² can be larger than 1.
  • Still report Cramér's V. Here it is .06, which tells the reader the observed association was very small.

APA 7 formatting rules

RuleCorrectIncorrect
Italicise statistical symbolsχ², N, p, V, φχ², N, p, V, φ (not italic)
Give df and N in the bracketsχ²(2, N = 150) = 13.79χ² = 13.79 or χ²(150) = 13.79
Report the exact p-value, to two or three decimalsp = .001p < .05
Very small p-values use "less than"p < .001p = .000
No leading zero for values that cannot exceed 1p = .001, V = .30p = 0.001, V = 0.30
Keep the leading zero on χ²χ² = 0.60χ² = .60
Round to two decimalsχ²(2, N = 150) = 13.79χ²(2, N = 150) = 13.786
Describe the pattern with percentages45.6% of part-time students preferred Both, compared with 18.3% of full-time studentsPart-time students preferred Both

SPSS prints "Sig. = .000" when p is very small. That never means p is zero, so report it as p < .001.

Common mistakes

  1. Skipping the expected count footnote. Check it, and switch to Fisher's exact test if too many cells are small.
  2. Leaving out N. APA 7 puts it in the brackets: χ²(2, N = 150).
  3. Reporting p without an effect size. Add Cramér's V, or φ for a 2 × 2 table, and say which it is.
  4. Stopping at χ². A significant result does not say where the association lies. Describe the percentages, and the residuals if you used them.
  5. Running the test on percentages or means. The test needs counts of people.
  6. Counting the same people twice. Each person belongs in one cell only. For before-and-after answers from the same people, use McNemar's test.
  7. Claiming a cause. An association does not show that one variable causes the other. Write "associated with", not "led to" or "caused".
  8. Calling a non-significant result "no relationship". Say there was insufficient evidence of an association.

You can report your result. But can you explain it?

Supervisors and examiners rarely ask you to recite the numbers. They ask what your result allows you to claim, and why you chose your test, effect size and interpretation. Could you answer these in a sentence, about your own result?

  1. “Your chi-square test is significant. Which categories drive the association, and how do you know?”
  2. “Why did you report Cramér's V rather than φ?”
  3. “Did your data meet the expected count assumption, and what would you have done if they had not?”

Defend This Result

This guide doesn't give model answers. In the Result Decoder, Defend This Result helps you practise answering questions like these about your own result, including the evidence that supports your conclusions and their important limitations. It can't predict the questions you'll be asked.

Practise with my own result

Frequently asked questions

Should I use Yates' continuity correction?

SPSS prints a Continuity Correction row only for 2 × 2 tables. Most guidance today reports the Pearson Chi-Square row, because the correction is overly cautious. If your expected counts are small, Fisher's exact test is the better choice. Whichever you use, name it in your write-up.

What if some expected counts are below 5?

Check the footnote under the Chi-Square Tests table. A common rule is that no more than 20% of cells should have an expected count below 5, and none below 1. If that fails, use Fisher's exact test (for a 2 × 2 table, or the Exact option for larger tables), or combine categories where that makes sense for your research question.

Should I report Cramér's V or phi?

Report phi for a 2 × 2 table and Cramér's V for anything larger. When one variable has only two categories, the two give the same number, but name the one that matches your table.

What if SPSS shows Sig. = .000?

Report p < .001. SPSS rounds very small p-values to .000, but the p-value is never exactly zero.

What is the difference between a chi-square test of independence and a goodness-of-fit test?

A test of independence looks at whether two categorical variables are associated. A goodness-of-fit test looks at one categorical variable and asks whether its counts match the proportions you expected.

Browse all Statistics Guides

Not sure this is the right test?

Answer a few questions about your study and the free Test Navigator suggests which test fits.

We use essential cookies to keep you signed in and the site working. With your permission, we also use optional analytics, including Google Analytics, to understand how Academic Decoder is used and which websites or social platforms bring visitors to it. Google may process this data in the United States. See our Privacy Policy.