Chi-Square Test: Goodness-of-Fit and Independence

What Is the Chi-Square Test?

The chi-square test is one of the most frequently encountered statistical tools in undergraduate and graduate coursework, and for good reason. Any time you are working with categorical data — data organized into groups or categories rather than measured on a continuous scale — the chi-square test is likely the right tool. It answers a deceptively simple question: do the frequencies you observed in your data match what you expected?

There are two major applications. The chi-square goodness-of-fit test compares an observed distribution against a theoretical or hypothesized one. The chi-square test of independence examines whether two categorical variables are related. Both tests share the same underlying logic and the same test statistic formula, but they are used in distinct research situations. If you need a refresher on how categorical versus continuous data differ before proceeding, the guide on the difference between qualitative and quantitative data is a helpful starting point.

Unlike the t-test or ANOVA — which compare means of continuous variables — the chi-square test is nonparametric. It makes no assumptions about the shape of the population distribution. This makes it particularly valuable in the social sciences, psychology, epidemiology, and marketing research, where many variables of interest are categorical by nature: gender, political affiliation, disease status, consumer preference, and so on.

Who Uses the Chi-Square Test?

Psychology researchers at institutions like Harvard University, Stanford University, and the University of Oxford routinely apply chi-square tests when comparing response frequencies across groups. Public health researchers at the Centers for Disease Control and Prevention (CDC) and the National Health Service (NHS) in the UK use it to test associations between exposures and disease outcomes in epidemiological surveys. Sociologists, political scientists, education researchers, and market researchers all rely on it when their dependent variable is categorical. According to a widely cited methodology textbook by Anderson, Sweeney, and Williams, chi-square tests for goodness-of-fit and independence are among the most commonly taught and applied statistical methods in business and economics courses in the United States and United Kingdom.

For students working on statistics assignments, this test appears in introductory statistics courses, research methods modules, biostatics programs, and advanced data analysis courses alike. Mastering it is not optional — it is foundational.

LSI and NLP Keywords Embedded Throughout This Guide

This article covers all the closely related concepts you need: observed frequencies, expected frequencies, degrees of freedom, p-value, null hypothesis, contingency table, Cramér's V, phi coefficient, Cohen's W, nonparametric test, categorical data, chi-square distribution, chi-square statistic, cross-tabulation, Fisher's exact test, Yates' correction, effect size, statistical significance, hypothesis testing, significance level, alpha level, SPSS chi-square, and chi-square in Excel.

The Chi-Square Distribution: What It Is and Why It Matters

Before you can interpret a chi-square test result, you need to understand the chi-square distribution itself. This is the probability distribution your test statistic follows when the null hypothesis is true. Understanding it prevents a very common student error: misinterpreting what a "significant" chi-square value actually means.

What Is the Chi-Square Distribution?

The chi-square distribution is a family of probability distributions, each defined by a single parameter: degrees of freedom (df). It is the distribution of the sum of squared standard normal random variables. When df is small, the distribution is strongly right-skewed. As df increases, it becomes more symmetric and begins to resemble a normal distribution. This matters for your chi-square test because the shape of the distribution — and therefore the critical values — depend entirely on degrees of freedom. You can look these values up in a chi-square distribution table or calculate them directly in software.

The chi-square distribution only takes non-negative values (0 and above). This makes sense: the chi-square statistic is computed from squared differences, which can never be negative. A chi-square value of 0 means the observed frequencies match expected frequencies perfectly. The further the chi-square statistic is from 0, the stronger the evidence against the null hypothesis.

How Degrees of Freedom Are Determined

Degrees of freedom determine which chi-square distribution you use to find your critical value or p-value. The formula differs between the two test types:

For example, a goodness-of-fit test with four categories has df = 3. A test of independence using a 3×4 contingency table has df = (3–1)(4–1) = 6. These degrees of freedom values go directly into your chi-square table or software when you look up your critical value or p-value. For a broader look at how hypothesis testing works across different tests, see the guide on hypothesis testing.

The Role of the P-Value

The p-value tells you the probability of obtaining a chi-square statistic as large as or larger than the one you calculated, assuming the null hypothesis is true. In most academic contexts — psychology, biology, sociology, business — the conventional alpha level is α = 0.05. If your p-value is less than 0.05, you reject the null hypothesis and conclude the result is statistically significant. If p > 0.05, you fail to reject the null hypothesis — you do not have enough evidence to conclude a significant difference or association exists.

Chi-Square Goodness-of-Fit Test: Definition, Formula, and Worked Example

The chi-square goodness-of-fit test tells you whether the distribution of a single categorical variable matches a hypothesized distribution. It is always a one-variable test. You have a set of categories, a set of observed counts, and a set of expected counts based on a theoretical model or prior knowledge. The test statistic measures how far the observed frequencies stray from the expected ones.

When to Use the Goodness-of-Fit Test

Use this test when you have one categorical variable and a specific hypothesis about how the observations should be distributed across its categories. Classic applications include:

• Testing whether die rolls from a six-sided die are equally distributed (each face appearing 1/6 of the time)
• Testing whether the racial composition of a university's student body matches the national population distribution
• Testing whether customer purchases are distributed evenly across days of the week
• Testing whether genetic inheritance follows Mendelian ratios (a foundational application in biology)
• Testing whether survey response categories match a known prior distribution

The expected frequencies can come from theory (like Mendelian genetics), a prior study, a population benchmark, or a simple assumption of equal distribution. The key point is that you must specify the expected distribution before looking at your data — not after. This connects directly to sound scientific method practice.

The Chi-Square Goodness-of-Fit Formula

This formula is the same for both the goodness-of-fit test and the test of independence. The difference lies in how you calculate the expected frequencies — and therefore in what the test is actually asking.

Assumptions of the Chi-Square Goodness-of-Fit Test

The chi-square test requires several conditions to be met before your results are trustworthy. Violating these assumptions can produce misleading conclusions. According to Laerd Statistics, the core assumptions are:

• One categorical variable: The variable must be measured at the nominal or ordinal level, organized into distinct categories.
• Random sampling: Your data must come from a random sample of the population of interest.
• Independence of observations: Each observation must be independent. One person's response must not influence another's.
• Mutually exclusive categories: Each observation falls into one and only one category — no overlap.
• Expected frequency rule: For two categories, each expected frequency must be at least 5. For three or more categories, each expected frequency must be at least 1, and no more than 20% of categories should have expected frequencies below 5.

If the expected frequency assumption is violated, the chi-square approximation becomes unreliable. In that situation, researchers turn to alternative approaches — particularly Fisher's exact test for small 2×2 tables, or collapsing categories where theoretically justifiable.

Worked Example: Goodness-of-Fit Test Step by Step

Research scenario: A professor at a large U.S. university believes that student absences are equally distributed across four categories: 0–2 absences, 3–5 absences, 6–8 absences, and 9+ absences. A random sample of 120 students produces the following observed counts:

Chi-Square Test of Independence: Definition, Contingency Tables, and Worked Example

The chi-square test of independence examines whether two categorical variables are related to each other — or whether they are statistically independent. This is the version of the chi-square test most commonly encountered in social science, psychology, and health research. The central question it answers: does knowing a person's category on variable X tell you anything about their likely category on variable Y?

What Is a Contingency Table?

A contingency table (also called a cross-tabulation or crosstab) is the data structure used in the chi-square test of independence. It displays the frequency distribution of two categorical variables simultaneously. Each cell in the table contains the count of observations that belong to that specific combination of categories. The row variable is typically the predictor or grouping variable; the column variable is typically the outcome or response variable.

For example, a 2×3 contingency table might display gender (male / female) across rows and preference for a product (strongly prefer / neutral / do not prefer) across columns. Each cell shows how many respondents fall into that gender-preference combination. Understanding how to read and construct a contingency table is a core skill in inferential statistics.

When to Use the Chi-Square Test of Independence

Use this test when you have two categorical variables and you want to test whether they are associated. Common research questions that call for it include:

• Is there an association between smoking status (smoker / non-smoker) and lung disease diagnosis (yes / no)?
• Does political party affiliation (Democrat / Republican / Independent) depend on educational attainment (high school / college / postgraduate)?
• Is brand preference associated with age group?
• Is there a relationship between a student's primary major and their choice of study method (alone / group / mixed)?
• Does treatment type (drug A / drug B / placebo) affect patient recovery category (full / partial / none)?

Calculating Expected Frequencies in a Contingency Table

In the test of independence, expected frequencies are not supplied by a theory. You calculate them from the marginal totals of your contingency table. The formula for each cell's expected frequency is:

This formula expresses what frequency you would expect in each cell if the two variables were completely independent — that is, if knowing a person's value on one variable told you nothing about their value on the other.

Worked Example: Test of Independence Step by Step

Research scenario: A researcher at a U.S. business school wants to know whether income level (Low / High) is associated with customer satisfaction (Satisfied / Dissatisfied) among 500 customers. The observed data are:

Assumptions of the Chi-Square Test and What to Do When They Are Violated

Many students compute a chi-square statistic without first checking whether the test is appropriate for their data. This is a significant methodological error. Every chi-square test — whether goodness-of-fit or independence — rests on a set of assumptions. When these assumptions are violated, your p-value is unreliable and your conclusions may be wrong. Checking assumptions is not optional procedure — it is the foundation of credible statistical reasoning, something emphasized in every rigorous research methods course from Columbia University to the London School of Economics.

The Expected Frequency Assumption in Depth

The expected frequency assumption is the one most frequently violated in practice — and the one that produces the most misleading results when ignored. The chi-square approximation works because, with sufficiently large samples, the sampling distribution of the test statistic follows a chi-square distribution. When expected frequencies are very small, this approximation breaks down. You cannot trust the p-value.

The conventional rule of thumb — originated by statistician William Cochran in his landmark 1952 paper — states that no more than 20% of expected frequencies should fall below 5, and none should be less than 1. When this is violated in a 2×2 table, use Fisher's exact test, which computes an exact p-value without relying on the chi-square approximation. When the problem affects a larger table, consider combining categories that have few observations — but only where that combination is theoretically defensible.

Yates' Continuity Correction

For 2×2 contingency tables with small expected frequencies (or when any expected frequency is between 5 and 25), some statisticians recommend applying Yates' continuity correction. This correction subtracts 0.5 from the absolute difference between each observed and expected frequency before squaring:

Yates' correction is controversial. Many statisticians argue it overcorrects, making the test too conservative. SPSS and R report both the Pearson chi-square and the Yates-corrected value for 2×2 tables so researchers can consider both. The safest approach in a 2×2 situation with small expected frequencies is Fisher's exact test. For guidance on running chi-square tests with real data, the walkthrough on how to run a chi-square test in SPSS covers this in detail.

Independence of Observations

The independence assumption is non-negotiable. If the same individuals provide data at two different time points — or if observations are paired or matched — the standard chi-square test is inappropriate. McNemar's test was developed precisely for this situation: it tests change in categorical responses for the same participants across two conditions. Pre-test/post-test designs, matched-pairs clinical trials, and repeated survey data all require McNemar's test rather than the standard chi-square.

Effect Size for Chi-Square Tests: Cohen's W, Phi, and Cramér's V

A statistically significant chi-square test only tells you that a difference or association is unlikely to be due to chance. It does not tell you how large or practically important that difference is. With large enough sample sizes, even trivially small associations become statistically significant. This is why reporting effect size is as important as reporting p-values — and why major publication guidelines from the American Psychological Association (APA) and the British Psychological Society (BPS) require it. For a broader treatment of effect size across statistical methods, see the guide on power analysis and Cohen's d.

Cohen's W — For Goodness-of-Fit Tests

Cohen's W is the standard effect size for the chi-square goodness-of-fit test. It is calculated as:

These benchmarks were proposed by statistician Jacob Cohen in his landmark 1988 book Statistical Power Analysis for the Behavioral Sciences. They are widely used across the social sciences, though Cohen himself emphasized that what constitutes a meaningful effect depends on the specific research domain.

Phi Coefficient (φ) — For 2×2 Contingency Tables

When your test of independence uses a 2×2 contingency table, the appropriate effect size measure is the phi coefficient. Phi (φ) is calculated from the chi-square statistic and ranges from 0 to 1:

Phi of 0.10 = small effect; 0.30 = medium effect; 0.50 = large effect. These mirror Cohen's W benchmarks and can be interpreted similarly.

Cramér's V — For Larger Contingency Tables

When your contingency table is larger than 2×2, phi overestimates the association because its maximum possible value exceeds 1. Cramér's V corrects for this by incorporating the table's dimensions:

Cramér's V = 0 means no association. V = 1 means a perfect association. The same Cohen benchmarks (0.10 / 0.30 / 0.50) apply as guidelines, though some researchers adjust these thresholds based on the degrees of freedom. Cramér's original 1946 paper established V as a standardized measure precisely because phi's limitation with larger tables was recognized even then.

Chi-Square Goodness-of-Fit vs. Test of Independence: Key Differences

Students frequently confuse these two tests, especially when writing up assignment methods sections. The tests share a formula and a distribution family, but they answer entirely different research questions. Getting this wrong in a methods section — using the wrong test name or describing the wrong research design — is a mark-losing error. The table below provides a clear comparison.

Related Tests You Should Know

The chi-square family is broader than just these two tests. Knowing the full landscape helps you choose appropriately in your research:

• Chi-square test of homogeneity: Tests whether the distribution of a categorical variable is the same across two or more independent populations. Often confused with the test of independence — both use contingency tables and the same formula, but homogeneity tests compare populations rather than examining association between variables.
• McNemar's test: A chi-square-based test for paired categorical data — used when the same participants are observed under two conditions.
• Cochran's Q test: Extends McNemar's test to three or more related groups. Used in repeated-measures designs with binary outcomes.
• Fisher's exact test: Exact probability calculation for 2×2 tables when expected frequencies are small. Does not rely on the chi-square approximation.
• Likelihood ratio chi-square (G-test): An alternative to Pearson's chi-square with similar properties. Preferred in some disciplines including ecology and linguistics.

Understanding which test applies when is a core competency in social statistics and research methods courses.

How to Run the Chi-Square Test in SPSS and Excel

Most statistics assignments and research projects at the university level now require software rather than hand calculation. The two platforms you are most likely to encounter are IBM SPSS Statistics (the standard in psychology, social science, and health research courses) and Microsoft Excel (ubiquitous in business and economics programs). Both can produce chi-square test results — but what they produce and how you interpret it differs in important ways.

Chi-Square Test in SPSS

SPSS produces the most complete chi-square output for the test of independence. For a full step-by-step walkthrough with screenshots, see the dedicated guide on running a chi-square test in SPSS. The key steps are:

• Navigate to Analyze → Descriptive Statistics → Crosstabs
• Place your row variable in the Row(s) box and your column variable in the Column(s) box
• Click Statistics and check "Chi-square," "Phi and Cramér's V," and "Contingency coefficient"
• Click Cells and check "Observed," "Expected," and "Row percentages" (or column, depending on your research question)
• Click OK and interpret the output in the Chi-Square Tests table

SPSS output for chi-square includes the Pearson Chi-Square, the Yates' Continuity Correction (for 2×2 tables), the Likelihood Ratio, degrees of freedom, and the asymptotic significance (two-tailed p-value). It also flags any cells with expected frequencies below 5 — SPSS automatically tells you whether your expected frequency assumption is met. This is the note printed at the bottom of the Chi-Square Tests output table: "a cells have expected count less than 5." If this note appears, you need to address the assumption violation.

Chi-Square Test in Excel

Excel's chi-square function — CHISQ.TEST(actual_range, expected_range) — returns the p-value directly from your observed and expected frequency arrays. It does not compute the test statistic for you, so you need to calculate χ² first using the standard formula across your cells. Excel is well-suited for the goodness-of-fit test when expected frequencies are pre-specified. For more complex contingency table analyses, SPSS or R offers more complete output. The guide on performing statistical tests in Excel illustrates the general workflow applicable across test types. For descriptive preparation before running chi-square, see the guide on statistical calculations in Excel.

Chi-Square in R and Python

For students and researchers using R, the base function chisq.test() handles both goodness-of-fit and independence tests. Feed it a vector of observed counts and (optionally) expected probabilities for goodness-of-fit, or a contingency table matrix for independence. The output includes the chi-square statistic, degrees of freedom, and p-value. The effectsize package provides Cramér's V directly. In Python, scipy.stats.chi2_contingency() computes the test of independence from a contingency table and returns the chi-square statistic, p-value, degrees of freedom, and expected frequency matrix in a single function call — making it easy to verify assumption compliance before interpreting results.

How to Report Chi-Square Results: APA Format and Academic Writing Standards

Knowing how to run a chi-square test is only half the job. Reporting it correctly in academic writing — using proper APA format, including the right elements, and interpreting results accurately — is what earns marks and meets journal publication standards. This section gives you the exact format and worked language for both test types.

APA Format for Chi-Square Goodness-of-Fit

APA Format for Chi-Square Test of Independence

Elements Every Chi-Square Report Must Include

• Test name: State explicitly whether you conducted a goodness-of-fit test or a test of independence
• Research question or hypothesis: Briefly state what you were testing
• Sample size (N): Include the total N in the chi-square notation
• Chi-square statistic: Reported to two decimal places with the symbol χ²
• Degrees of freedom: In parentheses immediately after χ²
• P-value: Report the exact value or "p < .001" for very small values
• Effect size: Cohen's W, phi, or Cramér's V as appropriate
• Interpretation: Describe the direction and nature of any significant finding in plain language
• Limitation note: For tests of independence, note that chi-square demonstrates association, not causation

For students working on full research papers, the research paper writing guide covers how to integrate statistical results sections into a broader academic argument. If you are struggling with the write-up structure itself, the literature review guide addresses how to frame quantitative findings within a broader literature context.

Common Chi-Square Mistakes and How to Fix Them

The chi-square test is conceptually straightforward — but students make the same errors repeatedly. Each of the following mistakes can undermine an otherwise solid statistical analysis. Recognizing them in your own work before submission is the difference between a strong grade and a disappointing one.

Mistake 1: Using Chi-Square with Continuous Data

The chi-square test is for categorical data — data organized into named groups. If your data are continuous (ages in years, test scores, income in dollars), the chi-square test is wrong. You would need correlation, regression, or a t-test depending on your research question. Artificially binning continuous data into categories to "make it work" for chi-square is poor practice — it discards information and reduces statistical power. The guide on qualitative vs. quantitative data is worth revisiting if you are unsure how to classify your variable.

Mistake 2: Ignoring the Expected Frequency Assumption

This is the most common technical error. Students compute a chi-square statistic and report results without checking whether expected frequencies meet the minimum threshold. When expected frequencies are too low, the chi-square distribution is a poor approximation of the actual sampling distribution. Always inspect your expected frequency matrix before interpreting results. SPSS flags this automatically. In manual calculations, compute every expected frequency before proceeding.

Mistake 3: Confusing Statistical Significance with Practical Importance

With a sample of 5,000, a trivially small association will be statistically significant at p < .001. This does not mean the finding is meaningful. A Cramér's V of 0.04 represents a negligible association regardless of the p-value. Always report and interpret effect size alongside your p-value. Large samples inflate chi-square statistics — understanding this prevents overclaiming. The guide on statistical power explains the relationship between sample size and significance in more depth.

Mistake 4: Inferring Causation from the Test of Independence

The chi-square test of independence establishes statistical association. It does not and cannot establish that one variable causes another. This is a foundational principle of statistical inference — and a common source of lost marks on methods and discussion sections. A significant χ² only tells you the two variables are not independent. Any causal claim requires additional experimental or longitudinal evidence, not a cross-sectional chi-square.

Mistake 5: Using Percentages Instead of Frequencies

The chi-square formula requires raw counts (frequencies), not percentages or proportions. Plugging percentages into the formula produces a meaningless chi-square value and an incorrect p-value. Always build your contingency table with actual observed counts. Report percentages in your interpretation section to describe patterns — but use counts in the calculation itself.

Mistake 6: Misidentifying the Test Type

Calling a test of independence a "goodness-of-fit test" in your methods section — or vice versa — signals conceptual confusion to your marker. These tests answer different questions, use different expected frequency calculations, and use different effect size measures. The comparison table in the previous section should help you keep them distinct.

Chi-Square Test Applications Across Academic Disciplines

One reason the chi-square test endures as a foundational tool is its extraordinary versatility. The same mathematical framework applies across disciplines that would otherwise share little methodological common ground. Understanding where and how it is used in your field makes the abstract formula concrete — and helps you recognize when it applies to your own research or assignment scenarios.

Psychology and Behavioral Science

In psychology, chi-square is used constantly. Clinical psychologists at institutions like King's College London and Johns Hopkins University use it to test whether treatment response categories (remission / partial response / no response) differ between therapy types. Social psychologists use the test of independence to examine whether implicit bias scores (categorized as high / medium / low) are associated with demographic variables. Educational psychologists test whether learning style categories are distributed differently across student populations from different socioeconomic backgrounds.

Epidemiology and Public Health

Public health researchers routinely use 2×2 contingency tables and the chi-square test of independence to assess associations between exposures and outcomes in observational studies. The CDC, the World Health Organization (WHO), and academic epidemiology departments at institutions like Johns Hopkins Bloomberg School of Public Health and University College London all publish findings derived from chi-square analyses. The measure of association in these contexts is often reported as a chi-square statistic alongside odds ratios and relative risks from logistic regression — for more on that relationship, see the logistic regression guide.

Genetics and Biology

The goodness-of-fit test has a particularly historic role in genetics. Gregor Mendel's laws of inheritance — dominant and recessive trait ratios like 3:1 and 9:3:3:1 — are tested using the chi-square goodness-of-fit test. When students in genetics labs at MIT, Cambridge, or any undergraduate biology program cross organisms and observe offspring ratios, they test whether their observed counts fit the Mendelian expected ratios using χ². This application predates the formal chi-square test — Mendel's data from the 1860s have been retrospectively analyzed using it. Understanding probability distributions more broadly supports this application — see the guide on probability distributions.

Business and Market Research

Market researchers use the chi-square test of independence to test whether consumer preferences differ across demographic groups. An advertising team might test whether product preference (brand A / B / C) is associated with age group (18–34 / 35–54 / 55+). Political pollsters test whether voting intention is independent of geographic region. Quality control engineers test whether defect categories are distributed equally across production lines. These practical applications appear in business statistics courses at schools like Wharton School, London Business School, and University of Chicago Booth School of Business. For a broader view of statistical methods applied to business problems, see the overview of regression analysis.

Sociology and Political Science

The chi-square test of independence is a workhorse of survey-based social research. Sociologists test whether attitudes toward immigration differ by income bracket. Political scientists test whether voter turnout categories differ between states with and without voter ID laws. The General Social Survey (GSS) administered by NORC at the University of Chicago and the British Social Attitudes Survey produce exactly the kind of categorical survey data that chi-square tests are designed to analyze. A representative published example: researchers examining survey data on social mobility test whether self-reported class identification (working / middle / upper) is independent of parental education level.

Frequently Asked Questions About the Chi-Square Test

Beyond Chi-Square: Related Statistical Tests for Categorical and Count Data

Mastering the chi-square test opens the door to a broader toolkit of methods for categorical and count data. Knowing when to move beyond chi-square — and what to use instead — is the mark of a statistically literate researcher. The following related tests are worth knowing for coursework and research across the disciplines where chi-square already applies.

Binomial Distribution and Exact Tests

When your categorical variable has exactly two outcomes (success / failure, yes / no, treated / control) and your sample is small, the binomial test is more appropriate than chi-square goodness-of-fit. It computes exact probabilities from the binomial distribution without relying on approximations. For a foundation in this distribution, the binomial distribution guide covers the relevant theory.

Multinomial Distribution

When your categorical variable has more than two categories and you are modeling frequencies across them, the multinomial distribution is the underlying probability model — of which the chi-square goodness-of-fit test is an approximate test. For a deeper treatment of multinomial models, see the multinomial distribution guide.

Logistic Regression

When your outcome is binary (yes / no) and you want to examine the effect of one or more predictors — including continuous ones — logistic regression is far more powerful than chi-square. It provides odds ratios, confidence intervals, and can control for confounding variables simultaneously. The logistic regression guide is essential reading for health, psychology, and social science researchers who move beyond chi-square. For understanding the assumptions that underpin regression models more broadly, the guide to regression assumptions is highly relevant.

T-Tests and ANOVA for Continuous Outcomes

If your research question involves comparing means of a continuous variable rather than frequencies of a categorical variable, you need a t-test (for two groups) or ANOVA (for three or more groups). These tests are covered separately but connect to chi-square through the shared logic of hypothesis testing and the calculation of test statistics, critical values, and p-values. For normal distribution theory underlying these tests, the guide on data distributions is a useful complement.

Type I and Type II Errors in Chi-Square Contexts

Every chi-square test — like every statistical test — carries the risk of two types of error. A Type I error occurs when you reject a true null hypothesis (a false positive). A Type II error occurs when you fail to reject a false null hypothesis (a false negative). The alpha level (typically 0.05) directly controls your Type I error rate. Statistical power — determined by sample size, effect size, and alpha — determines how well you avoid Type II errors. The comprehensive guide on Type I and Type II errors covers this in full for all hypothesis tests, including chi-square.

Confidence Intervals for Proportions

After a significant chi-square test, researchers often want to estimate the true proportion in each category with appropriate uncertainty quantification. Confidence intervals for proportions complement chi-square results by providing an estimate of the population proportion and the precision of that estimate. Reporting confidence intervals alongside p-values is now strongly encouraged by the APA and required by many peer-reviewed journals as part of good statistical reporting practice.