Chi-Square Test vs Fisher's Exact Test: When to Use Which

Use Fisher's exact test when expected cell counts fall below 5 in any cell of your contingency table, and use the chi-square test when all expected counts are 5 or higher. That single rule settles the most common decision researchers face when analyzing categorical data. When you are unsure which test fits your design, our statistical test selector points you to the right one. Karl Pearson introduced the chi-square test in 1900 as a goodness-of-fit measure, while Ronald Fisher developed the exact test in 1922 specifically for small-sample contingency tables. Over a century later, these two tests remain the foundation of categorical data analysis in clinical trials, epidemiological studies, and systematic reviews.

This guide walks you through every decision point, from hand calculations and continuity corrections to advanced stratified analyses, reporting standards, and R code you can run immediately.

Try our free Chi-Square Calculator to run either test instantly with automatic assumption checking.

Why Expected Cell Counts Drive the Decision

The chi-square test is an asymptotic approximation. It assumes that the sampling distribution of your test statistic follows a chi-square distribution, but this assumption only holds when sample sizes are large enough. The key metric is the expected cell frequency in each cell of your contingency table.

The rule: if any expected cell count is less than 5, switch to Fisher's exact test. To calculate expected counts, multiply each cell's row total by its column total, then divide by the grand total. For a 2x2 table with cells a, b, c, d and row totals R1, R2 and column totals C1, C2, the expected count for cell a equals (R1 x C1) / N.

When expected counts are borderline (between 5 and 10), both tests typically produce similar p-values. The divergence becomes meaningful when expected counts drop below 5, because the chi-square approximation inflates the Type I error rate in that range. Some textbooks use a threshold of 10 instead of 5, but the widely accepted convention in biomedical research is 5.

Step-by-Step Calculation Example with Real Clinical Data

Consider a randomized controlled trial comparing a new antiplatelet drug to placebo for preventing stroke. The 2x2 contingency table:

Step 1: Calculate expected cell counts.

• Expected (Drug, Stroke) = (100 x 36) / 200 = 18.0
• Expected (Drug, No Stroke) = (100 x 164) / 200 = 82.0
• Expected (Placebo, Stroke) = (100 x 36) / 200 = 18.0
• Expected (Placebo, No Stroke) = (100 x 164) / 200 = 82.0

All expected counts exceed 5, so the chi-square test is appropriate here.

Step 2: Compute the chi-square statistic.

For each cell, calculate (Observed - Expected)^2 / Expected:

• (12 - 18)^2 / 18 = 36 / 18 = 2.00
• (88 - 82)^2 / 82 = 36 / 82 = 0.439
• (24 - 18)^2 / 18 = 36 / 18 = 2.00
• (76 - 82)^2 / 82 = 36 / 82 = 0.439

Chi-square statistic = 2.00 + 0.439 + 2.00 + 0.439 = 4.878

Step 3: Determine the p-value.

With 1 degree of freedom (df = (rows - 1) x (cols - 1) = 1), the p-value for a chi-square of 4.878 is approximately 0.027. This is below the conventional alpha of 0.05, suggesting a statistically significant association between drug assignment and stroke occurrence.

Step 4: Calculate the effect size.

The phi coefficient = sqrt(4.878 / 200) = sqrt(0.02439) = 0.156, indicating a small-to-medium effect. You can compute this instantly using our effect size estimator.

Now imagine the same proportions but with a total sample of only 20 patients (Drug: 1 stroke, 9 no stroke; Placebo: 3 strokes, 7 no stroke). The expected count for the Drug/Stroke cell would be (10 x 4) / 20 = 2.0, which falls below 5. In that scenario, you would use Fisher's exact test instead.

Fisher's Exact Test

Fisher's exact test calculates the exact probability of observing your data (or a more extreme result) given the null hypothesis of no association. It does not rely on an approximation, which is why it remains valid regardless of sample size. Fisher developed this approach using the hypergeometric distribution, which models the probability of each possible 2x2 table configuration given fixed row and column totals.

For the small-sample stroke example above (1, 9, 3, 7 with N = 20), Fisher's exact test would enumerate all possible tables with those marginals and sum the probabilities of tables as extreme or more extreme than the observed one. The resulting two-sided p-value would be approximately 0.582, indicating no significant association. This contrasts sharply with what a naive chi-square application would suggest, demonstrating why test selection matters.

Fisher's exact test is primarily designed for 2x2 tables. For larger tables, the Freeman-Halton test generalizes the logic, though computation becomes intensive as table dimensions grow.

Chi-Square with Yates' Continuity Correction

Frank Yates introduced his continuity correction in 1934 to bridge the gap between the discrete chi-square test statistic and the continuous chi-square distribution. The correction subtracts 0.5 from the absolute difference between observed and expected counts before squaring:

Corrected chi-square = Sum of (|Observed - Expected| - 0.5)^2 / Expected

For our stroke example, the corrected statistic would be:

• (|12 - 18| - 0.5)^2 / 18 = (5.5)^2 / 18 = 1.681
• And similarly for the other three cells.

The corrected chi-square = 4.167, yielding p = 0.041 (compared to p = 0.027 without correction).

When to apply Yates' correction:

• Only for 2x2 tables (never for larger contingency tables)
• When expected counts are between 5 and 10 in one or more cells
• When you want a conservative estimate that approximates the two-sided Fisher's exact p-value

When to skip it:

• When all expected counts comfortably exceed 10
• When you already plan to use Fisher's exact test for small samples
• When journal guidelines or reviewers specifically request the uncorrected version

Many modern statisticians, including those publishing simulation studies in Statistics in Medicine and the British Medical Journal, now consider Yates' correction overly conservative for routine use. It reduces statistical power, potentially masking genuine effects. The safest approach: if expected counts are borderline, skip the correction and run Fisher's exact test directly.

McNemar's Test for Paired and Matched Data

Neither the chi-square test nor Fisher's exact test is appropriate when your categorical observations are paired or matched. Consider a before-and-after study where the same 80 patients are tested for antibiotic resistance before and after treatment:

The standard chi-square test would treat these 80 paired observations as if they were 80 independent observations, violating the independence assumption. McNemar's test focuses only on the discordant pairs (cells b and c), where change actually occurred.

McNemar's test statistic = (b - c)^2 / (b + c) = (25 - 5)^2 / (25 + 5) = 400 / 30 = 13.33

With 1 degree of freedom, this yields p < 0.001, indicating a significant change in resistance patterns after treatment.

When the number of discordant pairs (b + c) is small (below 25), use the exact binomial version of McNemar's test instead of the chi-square approximation.

Common scenarios requiring McNemar's test in clinical research include pre-post intervention studies, matched case-control designs, and diagnostic test comparisons where the same patients undergo two different tests.

Cochran-Mantel-Haenszel Test for Stratified Tables

When you need to test an association while controlling for a confounding variable, the Cochran-Mantel-Haenszel (CMH) test is the standard approach. William Cochran proposed the method in 1954, and Nathan Mantel and William Haenszel extended it in 1959 for case-control studies.

Imagine testing whether a drug reduces infection across three hospitals. Each hospital produces a separate 2x2 table, and infection rates differ by hospital (a confounding variable). The CMH test combines evidence across all strata (hospitals) to produce a single pooled test of association while adjusting for the stratifying variable.

The CMH statistic follows a chi-square distribution with 1 degree of freedom under the null hypothesis of no association across strata. It also produces a common odds ratio (the Mantel-Haenszel odds ratio), which is the weighted average of stratum-specific odds ratios.

When to use CMH:

• Multicenter clinical trials with center as a stratifying variable
• Studies adjusting for a categorical confounder (age group, sex, disease severity)
• Systematic reviews combining 2x2 tables across studies (this is essentially what a fixed-effect meta-analysis does with the Mantel-Haenszel method)

The CMH test assumes a consistent direction of association across strata (no qualitative interaction). Test this assumption with the Breslow-Day test for homogeneity of odds ratios before interpreting the CMH result.

For researchers conducting systematic reviews and meta-analyses, understanding the CMH framework is essential because the Mantel-Haenszel method is one of the two primary approaches (alongside inverse-variance) for pooling binary outcome data. Learn more about choosing the right statistical test for your research.

Larger Contingency Tables: 3x2, 4x2, and Beyond

The chi-square test generalizes naturally to tables larger than 2x2. A 3x2 table (3 exposure groups, 2 outcomes) has (3 - 1) x (2 - 1) = 2 degrees of freedom. A 4x3 table has 6 degrees of freedom.

For these larger tables, the expected count rule still applies: if more than 20% of cells have expected counts below 5, or if any cell has an expected count below 1, the chi-square approximation becomes unreliable.

Fisher's exact test becomes computationally impractical for larger tables because the number of possible configurations grows exponentially. A 2x2 table with N = 20 might have a few dozen possible configurations. A 4x5 table with N = 200 could have millions. For these situations, Monte Carlo simulation provides the solution: randomly generate thousands (typically 10,000 or more) of tables with the same marginals, compute the test statistic for each, and estimate the p-value from the empirical distribution.

Practical guidance for larger tables:

When your variable is ordinal (e.g., disease severity: mild, moderate, severe), a chi-square test for linear trend (also called the Cochran-Armitage test) is more powerful than the standard chi-square test because it exploits the ordering information.

Harald Cramer introduced Cramer's V in 1946 as the standard effect size measure for tables larger than 2x2. For a table with r rows and c columns, Cramer's V = sqrt(chi-square / (N x min(r - 1, c - 1))). Benchmarks for Cramer's V depend on the smaller dimension: for df* = 1, small = 0.10, medium = 0.30, large = 0.50; for df* = 2, small = 0.07, medium = 0.21, large = 0.35.

Exact vs. Asymptotic P-Values: What the Difference Means Clinically

The distinction between exact p-values and asymptotic p-values often confuses researchers, but it has real clinical implications.

An asymptotic p-value comes from the chi-square distribution, which is a theoretical distribution that the test statistic approaches as sample size grows toward infinity. For large samples, this approximation is excellent. For small samples, it can be inaccurate in either direction.

An exact p-value is calculated by enumerating all possible data configurations and summing their probabilities. It makes no distributional assumptions and is always valid, regardless of sample size.

When do they diverge clinically?

Consider a small pilot study (N = 30) that produces a chi-square p-value of 0.042 and a Fisher's exact p-value of 0.071. The chi-square result would lead you to reject the null hypothesis, while Fisher's exact test would not. This discrepancy could determine whether a pilot study proceeds to a full-scale trial, influencing thousands of future patients.

In regulatory submissions and clinical trial reports, the FDA and EMA both accept Fisher's exact test for small samples. Many journal guidelines (including CONSORT for randomized trials) recommend reporting the exact p-value when sample sizes are small. If both tests agree (both significant or both non-significant), report either one. If they disagree, report Fisher's exact test and note the discrepancy.

For a deeper understanding of p-values in the context of research synthesis, see our guide on p-values and confidence intervals explained.

How to Report Chi-Square and Fisher's Results in a Manuscript

Clear reporting is essential for reproducibility. Follow these APA-style reporting formats:

Chi-square test: "A chi-square test of independence was performed to examine the relationship between treatment group and stroke occurrence. The association was statistically significant, chi-square(1, N = 200) = 4.88, p = .027, phi = 0.16."

Chi-square with Yates' correction: "With Yates' continuity correction, the association remained significant, chi-square(1, N = 200) = 4.17, p = .041, phi = 0.16."

Fisher's exact test: "Because two cells had expected counts below 5, Fisher's exact test was used. The association between treatment and outcome was not statistically significant (p = .582, two-tailed)."

McNemar's test: "McNemar's test indicated a significant change in antibiotic resistance after treatment, chi-square(1) = 13.33, p < .001."

Key reporting elements to include every time:

• The test name and why it was chosen (especially for Fisher's exact test)
• Degrees of freedom and sample size for chi-square
• The exact p-value (not just "p < 0.05")
• An effect size measure (phi for 2x2, Cramer's V for larger tables)
• Whether the test was one-tailed or two-tailed

Reviewers frequently reject manuscripts that report chi-square results without effect sizes. Always pair your statistical significance with a measure of practical significance.

Need help with the statistical analysis section of your manuscript? Our biostatisticians handle everything from test selection to APA-formatted results tables. Request a free quote and tell us about your project.

Using Chi-Square Results in Meta-Analysis

When conducting a meta-analysis of studies reporting binary outcomes, you often encounter studies that report only a chi-square statistic or a p-value rather than a full 2x2 table. Converting these values to effect sizes suitable for pooling requires a few steps.

Converting chi-square to an odds ratio:

For a 2x2 table, you can recover the log odds ratio from the chi-square statistic if you also know the marginal totals. If only the chi-square and N are available, you can approximate:

• phi = sqrt(chi-square / N)
• For a 2x2 table with balanced groups: OR is approximately exp(phi x pi / sqrt(3))

However, this approximation is rough. Whenever possible, extract or reconstruct the full 2x2 table from the original paper. Contact study authors if the table is not reported.

Converting a p-value to a test statistic:

If a study reports only "p = 0.03" for a chi-square test with 1 df, you can recover the chi-square statistic using the inverse chi-square distribution. In R: qchisq(1 - 0.03, df = 1) gives approximately 4.71.

Why this matters for systematic reviews:

In a systematic review, inconsistent test reporting across primary studies is one of the biggest obstacles to quantitative synthesis. Some studies report Fisher's exact p-values, others report chi-square statistics, and some report only odds ratios with confidence intervals. Understanding the relationship between these quantities lets you convert between formats and include more studies in your meta-analysis, increasing statistical power.

For converting between effect size formats, use our Cohen's d calculator.

R Code for Both Tests

Here is reproducible R code for running both the chi-square test and Fisher's exact test, along with Yates' correction, McNemar's test, and the CMH test.

# === Chi-Square Test vs Fisher's Exact Test in R ===

# Create the 2x2 contingency table (stroke example)
stroke_table <- matrix(c(12, 88, 24, 76), nrow = 2, byrow = TRUE,
                       dimnames = list(Group = c("Drug", "Placebo"),
                                       Outcome = c("Stroke", "No Stroke")))
print(stroke_table)

# Check expected cell counts
expected <- chisq.test(stroke_table)$expected
print(expected)  # All should be >= 5 for chi-square

# --- Chi-square test (without Yates' correction) ---
chi_result <- chisq.test(stroke_table, correct = FALSE)
print(chi_result)
# Extract: chi_result$statistic, chi_result$p.value

# --- Chi-square test (with Yates' continuity correction) ---
chi_yates <- chisq.test(stroke_table, correct = TRUE)
print(chi_yates)

# --- Fisher's exact test ---
fisher_result <- fisher.test(stroke_table)
print(fisher_result)
# Provides: p-value, odds ratio, 95% CI for odds ratio

# --- Effect size: Phi coefficient ---
n <- sum(stroke_table)
phi <- sqrt(chi_result$statistic / n)
cat("Phi coefficient:", round(phi, 3), "\n")

# --- Effect size: Cramer's V (for larger tables) ---
# For a table with r rows and c columns:
# cramers_v <- sqrt(chi_result$statistic / (n * (min(nrow(table), ncol(table)) - 1)))

# --- McNemar's test (paired data) ---
paired_table <- matrix(c(10, 25, 5, 40), nrow = 2, byrow = TRUE,
                       dimnames = list(Before = c("Resistant", "Susceptible"),
                                       After = c("Resistant", "Susceptible")))
mcnemar_result <- mcnemar.test(paired_table)
print(mcnemar_result)

# Exact McNemar (when discordant pairs < 25)
mcnemar_exact <- binom.test(paired_table[1,2], paired_table[1,2] + paired_table[2,1])
print(mcnemar_exact)

# --- Cochran-Mantel-Haenszel test (stratified data) ---
# Example: drug vs placebo across 3 hospitals
hospital1 <- matrix(c(5, 45, 10, 40), nrow = 2)
hospital2 <- matrix(c(3, 47, 8, 42), nrow = 2)
hospital3 <- matrix(c(4, 46, 6, 44), nrow = 2)
stratified <- array(c(hospital1, hospital2, hospital3), dim = c(2, 2, 3),
                    dimnames = list(Group = c("Drug", "Placebo"),
                                    Outcome = c("Event", "No Event"),
                                    Hospital = c("H1", "H2", "H3")))
cmh_result <- mantelhaen.test(stratified)
print(cmh_result)
# Provides: chi-square statistic, p-value, common odds ratio, 95% CI

# --- Fisher's test for larger tables (Freeman-Halton) ---
large_table <- matrix(c(10, 20, 5, 15, 25, 10, 8, 30, 12), nrow = 3)
fisher_large <- fisher.test(large_table)
print(fisher_large)

# --- Monte Carlo simulation (for very large tables) ---
fisher_mc <- fisher.test(large_table, simulate.p.value = TRUE, B = 10000)
print(fisher_mc)

By default, R's chisq.test() applies Yates' correction for 2x2 tables. Set correct = FALSE to disable it. Fisher's fisher.test() always returns a two-sided p-value by default; specify alternative = "greater" or alternative = "less" for one-sided tests.

When Reviewers Ask You to Switch Tests

Peer reviewers frequently request test changes during manuscript revision. Here is how to handle the most common scenarios. If the statistical back-and-forth is stalling your acceptance, you can bring in a PhD statistician to rerun the analysis and draft the response.

"Please use Fisher's exact test instead of chi-square."

This is the most common request. If your expected counts were all above 5, you can explain that the chi-square test was appropriate, but running Fisher's exact test as well costs nothing. Report both results in a supplementary note. For large samples, the p-values will be nearly identical, which strengthens your analysis rather than undermining it.

"Please use chi-square instead of Fisher's exact test."

This request sometimes comes from reviewers who believe Fisher's exact test is only for small samples. Fisher's exact test is valid for any sample size, so using it is never incorrect. Explain this in your response, cite the relevant statistical literature, and offer to include the chi-square result alongside.

"Your chi-square test is inappropriate because your data are paired."

If the reviewer is correct, you need to switch to McNemar's test. This is a substantive change that could alter your conclusions. Reanalyze and report the corrected results transparently.

"Please adjust for [confounding variable]."

This calls for the Cochran-Mantel-Haenszel test or logistic regression, depending on the nature of the confounder. Stratify your data and run the CMH test if the confounder is categorical with a small number of levels.

"Please report effect sizes."

Calculate phi (2x2) or Cramer's V (larger tables) and add them to your results. This is a reasonable request that improves reporting quality. See our Hedges' g calculator for quick computation.

For help navigating reviewer comments on your statistical analysis, our biostatistics service includes a dedicated revision support package.

Decision Table

Effect Sizes: Cramer's V and Phi Coefficient

For 2x2 tables, use the phi coefficient (ranges from 0 to 1). Benchmarks following Cohen's conventions: small = 0.10, medium = 0.30, large = 0.50.

For larger tables, use Cramer's V, which adjusts phi by table dimensions. The formula is V = sqrt(chi-square / (N x (k - 1))), where k is the smaller of the number of rows or columns.

Always report effect size alongside your p-value. A significant chi-square result with a Cramer's V of 0.05 suggests a trivially small association that may have no practical importance despite statistical significance. Conversely, a non-significant result with a moderate effect size may indicate that your study was underpowered rather than that no association exists.

For diagnostic test accuracy data, see our Diagnostic Accuracy Calculator.

Applying This in Systematic Reviews and Meta-Analyses

In systematic reviews, the chi-square test and Fisher's exact test appear in two distinct contexts.

First, as reported statistics in primary studies. During data extraction, check whether original study authors chose the appropriate test given their sample sizes. Flag studies that applied chi-square with inadequate expected cell counts, as this introduces potential error into the evidence base. Record the exact p-values and, when available, the full 2x2 tables for later meta-analytic pooling.

Second, as part of the meta-analytic framework itself. Cochran's Q test for heterogeneity is fundamentally a chi-square test that evaluates whether variation in effect sizes across studies exceeds what chance alone would produce. Similarly, the Mantel-Haenszel method for pooling odds ratios across studies is a direct extension of the CMH test discussed earlier in this guide.

When pooling binary outcomes, the choice between the Mantel-Haenszel method and the inverse-variance method depends on the number of studies and the frequency of zero events. The Mantel-Haenszel approach handles zero cells more gracefully and is preferred when studies have small sample sizes or rare events.

For inter-rater agreement on categorical outcomes during screening or data extraction, the Kappa Calculator handles Cohen's kappa and provides interpretation benchmarks.

Common Mistakes

Checking total N instead of expected counts. A study with N = 200 can still have cells with expected counts below 5 if the outcome is rare. Always compute expected counts cell by cell.

Applying chi-square to paired data. This violates the independence assumption and will produce misleading p-values. Use McNemar's test instead.

Forgetting to report effect sizes. A p-value without phi or Cramer's V tells the reader nothing about practical significance. Many journals now require effect sizes as part of their reporting guidelines.

Assuming Fisher's exact test is always more conservative. For large samples, both tests give virtually identical p-values. For very small samples, Fisher's can actually produce smaller p-values than the chi-square test in certain configurations.

Using Yates' correction reflexively. Apply it only for 2x2 tables with borderline expected counts, and consider Fisher's exact test as a better alternative.

Ignoring the Breslow-Day test before interpreting a CMH result. If odds ratios vary substantially across strata (significant Breslow-Day test), the pooled CMH odds ratio is misleading.

Reporting one-tailed Fisher's exact p-values without justification. The default should always be two-tailed unless you have a strong a priori directional hypothesis stated before data collection.

Key Takeaways

• Use Fisher's exact test when any expected cell count is below 5 or when total N is below 20.
• Use the chi-square test for larger samples where all expected cell counts meet the threshold.
• Yates' continuity correction bridges discrete and continuous distributions but is often overly conservative. Use Fisher's exact test instead for small samples.
• For paired categorical data, use McNemar's test, never chi-square or Fisher's.
• For stratified analyses adjusting for confounders, use the Cochran-Mantel-Haenszel test.
• For larger tables with small expected counts, use Freeman-Halton or Monte Carlo simulation.
• Always report phi (2x2 tables) or Cramer's V (larger tables) as your effect size measure.
• When converting chi-square results for meta-analysis, reconstruct full 2x2 tables whenever possible.
• Exact and asymptotic p-values can diverge meaningfully in small samples, potentially changing clinical conclusions.
• Statistical significance and effect size together tell the complete story.