What Is a Proportion Test (One- and Two-Sample z-Test for Proportions)?

A proportion test is a statistical hypothesis test used to determine whether the observed proportion (percentage) in a sample differs significantly from a known or hypothesized population proportion—or whether two independent sample proportions differ from each other. It is the go-to tool for quality engineers working with defect rates, pass/fail data, or any binary outcome.

---

What It Is

In quality management, we often deal with counts: how many units failed, how many customers complained, how many parts were nonconforming. A proportion test answers questions like:

• Is our current defect rate significantly lower than the historical 5% target
• Did Process A produce a significantly different yield than Process B

The test compares the observed sample proportion (p̂) against a reference value (p₀) or against another sample proportion, while accounting for sampling variability.

---

How It Works: Formula and Steps

### One-Sample Proportion Test (z-test)

Hypotheses:

• Null hypothesis (H₀): p = p₀ (the true proportion equals the hypothesized value)
• Alternative hypothesis (H₁): p ≠ p₀, p > p₀, or p < p₀ (two-tailed or one-tailed)

Test statistic:

\[
z = \frac{\hat{p} - p_0}{\sqrt{\frac{p_0(1 - p_0)}{n}}}
\]

Where:

• p̂ = x / n (sample proportion, with x = number of successes)
• p₀ = hypothesized population proportion
• n = sample size

Decision rule:
Compare the computed z-value to the critical value from the standard normal distribution. For a 95% confidence level (α = 0.05, two-tailed), the critical value is z = 1.96. Reject H₀ if |z| > 1.96.

### Two-Sample Proportion Test

Hypotheses:

• H₀: p₁ = p₂
• H₁: p₁ ≠ p₂ (or one-tailed)

Test statistic:

\[
z = \frac{\hat{p}_1 - \hat{p}_2}{\sqrt{\hat{p}(1 - \hat{p})\left(\frac{1}{n_1} + \frac{1}{n_2}\right)}}
\]

Where p̂ = (x₁ + x₂) / (n₁ + n₂) is the pooled proportion.

Assumptions:

• Random, independent samples
• Sample size large enough so that np ≥ 5 and n(1−p) ≥ 5 (normal approximation holds)

---

A Worked Illustrative Example

Example data (illustrative only):
A factory claims its defect rate is at most 4% (p₀ = 0.04). A quality audit inspects n = 400 units and finds x = 24 defects, giving p̂ = 24/400 = 0.06.

Step 1 – Hypotheses:
H₀: p = 0.04 vs. H₁: p > 0.04 (one-tailed test)

Step 2 – Compute z:

\[
z = \frac{0.06 - 0.04}{\sqrt{\frac{0.04 \times 0.96}{400}}} = \frac{0.02}{\sqrt{0.000096}} = \frac{0.02}{0.0098} \approx 2.04
\]

Step 3 – Compare to critical value:
For a one-tailed test at α = 0.05, the critical z is 1.645. Since 2.04 > 1.645, we reject H₀.

Conclusion: There is statistically significant evidence that the true defect rate exceeds the claimed 4%. (Note: this is illustrative data, not real factory results.)

---

Common Pitfalls

1. Small sample sizes: The normal approximation fails when np or n(1−p) is below 5. Use an exact binomial test instead.
2. Ignoring the direction of the test: A one-tailed test requires a different critical value than a two-tailed test.
3. Using the sample proportion in the denominator: For the one-sample test, always use the hypothesized p₀ in the standard error, not p̂.
4. Confusing statistical significance with practical importance: A large sample can make a tiny difference "significant" even when it is not meaningful for the business.

---

Try It Yourself

Manual calculation is fine for a single test, but in daily quality work you will run many comparisons. Use the free, no-sign-up proportion test calculator at https://www.6sq.com/tools/proportion_test/ to get instant z-values and p-values—just enter your sample counts and hypothesized proportion.