Why Ordered Proportion Testing Matters
Many studies compare a binary outcome across ordered groups. The groups may be dose levels, exposure bands, age classes, stages, or ranked scores. A simple chi square test can show that proportions differ. It does not focus on a steady ordered rise or fall. The Cochrane Armitage trend test targets that ordered pattern.
What The Test Measures
The method compares observed successes with expected successes under no trend. Each group receives a numeric score. Those scores define the order and spacing. The calculator centers the scores by the group totals. It then builds a z statistic from weighted deviations. A large positive z suggests increasing risk. A large negative z suggests decreasing risk.
Good Data Preparation
Use counts, not percentages. Enter the number of successes and the group total for every ordered category. The total must include successes and non successes. Keep the order meaningful. Low dose should appear before high dose. Early stage should appear before late stage. Scores can be simple ranks, such as 1, 2, 3, and 4. Unequal scores are useful when the categories have real distances.
Interpreting Results
The two sided p value checks for any linear trend. The greater p value checks an upward trend. The less p value checks a downward trend. Select the alternative before reading the result. Also review the group rates. A significant test can still hide one unusual group. Small counts may need exact or permutation methods.
Practical Uses
Researchers use this test in clinical trials, toxicology, screening studies, survey analysis, and quality control. It is helpful when the question is ordered by design. It is less helpful when categories have no natural rank. Always explain the scoring choice in reports. Clear scores make the result easier to audit. Use subject expertise when the trend size appears modest or uncertain overall.
Limits And Assumptions
The test assumes independent observations and binomial outcomes within groups. It also expects the chosen scores to represent the planned order. It is a large sample method, so very sparse tables can be unstable. Use it with charts, rates, and context. Statistical significance is not the same as practical importance. Report both the trend test and the observed proportions together.