What Is Hypothesis Testing and How Does It Improve Quality Decisions?

In quality engineering, we constantly face decisions: Is this process running at its target? Did a new supplier actually reduce defect rates? Hypothesis testing—a cornerstone of statistical quality control—gives you a disciplined, data-driven way to answer such questions without relying on gut feel.

What It Is

Hypothesis testing is a formal procedure, rooted in the Neyman-Pearson framework, for deciding whether a claim about a population parameter (mean, variance, proportion, etc.) is supported by sample data. You start with two competing statements:

• Null hypothesis (H₀): The default assumption—typically "no change," "no effect," or "status quo." For example, the process mean equals the target value.
• Alternative hypothesis (H₁ or Ha): What you want to prove—e.g., the mean has shifted, or the new process is better.

The goal is to gather evidence against H₀. If the evidence is strong enough, you reject H₀ in favor of H₁; otherwise, you fail to reject H₀ (you do not "accept" H₀—you simply lack proof against it).

How It Works: Key Steps and Concepts

1. State the hypotheses. Define H₀ and H₁ clearly, in terms of population parameters (e.g., H₀: μ = μ₀ vs. H₁: μ ≠ μ₀ for a two-sided test, or μ > μ₀ for a one-sided test).

1. Choose the significance level (α). This is the risk of rejecting H₀ when it is actually true (Type I error). Common choices are α = 0.05 or 0.01.

1. Select the test statistic. The choice depends on your data and what you are testing:

- z-test – for means when the population standard deviation is known (or sample size is large, n ≥ 30).
- t-test – for means when σ is unknown (use sample standard deviation s).
- F-test – for comparing two variances.
- χ²-test – for goodness-of-fit or tests of independence.

1. Compute the p-value. The p-value is the probability of observing a test statistic as extreme as, or more extreme than, the one computed from your sample, assuming H₀ is true.

1. Make a decision.

- If p-value ≤ α, reject H₀ (result is "statistically significant").
- If p-value > α, fail to reject H₀.

For a z-test on a mean, the test statistic is:

z = (x̄ − μ₀) / (σ / √n)

where x̄ is the sample mean, μ₀ is the hypothesized mean, σ is the population standard deviation, and n is the sample size. For a t-test, replace σ with s and use the t-distribution with n−1 degrees of freedom.

A Worked Illustrative Example

Example data (illustrative only): A machine is supposed to fill bags with 500 g of product. You take a random sample of n = 25 bags and find a sample mean x̄ = 505 g, with sample standard deviation s = 15 g. You want to test whether the machine is off-target, using α = 0.05.

• Step 1: H₀: μ = 500 g; H₁: μ ≠ 500 g (two-sided test).
• Step 2: α = 0.05.
• Step 3: Since σ is unknown, use a t-test. Test statistic:

t = (505 − 500) / (15 / √25) = 5 / 3 = 1.667

• Step 4: Degrees of freedom = n − 1 = 24. For a two-tailed test with df = 24, the critical t-value at α = 0.05 is approximately 2.064. The p-value for t = 1.667 is approximately 0.108.
• Step 5: Since p-value (0.108) > α (0.05), you fail to reject H₀. There is not enough evidence to conclude the machine is off-target.

Common Pitfalls

• Confusing "statistical significance" with "practical importance." A very large sample can make a tiny, meaningless difference statistically significant.
• Misinterpreting the p-value. It is not the probability that H₀ is true; it is the probability of your data (or more extreme) given that H₀ is true.
• Choosing a one-sided test after seeing the data. Decide the direction before collecting data to avoid bias.
• Ignoring assumptions. t-tests assume approximate normality; F-tests assume normality and independence. Check your data first.

---

Hypothesis testing turns raw sample data into a clear, defensible "go/no-go" decision for process improvements, supplier evaluations, and quality audits. To run these calculations quickly and avoid manual errors, try the free, interactive tool at 6SQ Hypothesis Testing.