Z-Test for Proportions

Test hypotheses about population proportions using the normal approximation.

Test Overview

What it tests:

Compares observed proportions to expected proportions or compares proportions between two groups using normal approximation to binomial distribution for hypothesis testing.

When to Use:

• • Binary outcomes (success/failure)
• • Large sample sizes meeting criteria
• • Compare proportions to standards
• • Test specific proportion values
• • Quality control applications

Key Assumptions:

• • np ≥ 5 and n(1-p) ≥ 5
• • Independent observations
• • Random sampling from population
• • Fixed sample size determined in advance
• • Binomial distribution conditions met

Test Statistic & Formula Focus

One-sample Z-test for a proportion

= observed sample proportion

= hypothesized population proportion

= sample size

Two-sample Z-test for comparing proportions

ˆp1,ˆp2

= observed proportions for group 1 and 2

= pooled proportion across both groups

n1,n2

= sample sizes for group 1 and 2

Reading the Proportion Comparison Plot

Example Result & Interpretation

Sample Size Requirements:

Normal approximation is valid when both np ≥ 5 and n(1-p) ≥ 5 for expected counts. When these conditions aren't met, use exact binomial tests instead.

Common Applications:

Quality Control:

• • Defect rates vs. specifications
• • Process capability assessment
• • Compliance testing
• • Acceptance sampling

Research Applications:

• • Clinical trial success rates
• • Marketing conversion rates
• • Polling and survey analysis
• • A/B testing comparisons

sixsigmacenter.com Implementation Guide

Step-by-Step Calculation:

The platform automates proportion calculations, pooled variance for two-sample tests, and reports p-values with confidence intervals for immediate business interpretation.

Example Data Layout for Z-Test for Proportions

Count data showing number of successes and total trials for each group or sample. Use for testing proportions/percentages. Sample Size Recommendation: Both np ≥ 5 and n(1-p) ≥ 5 must be satisfied for normal approximation validity - typically requires n ≥ 30 for adequate test performance.

This is how your spreadsheet or CSV file should be structured before you upload it. Match the column names, keep one row per observation, and leave the analysis to the platform.