Cochran-Armitage Trend Test
Tests whether the rate of a binary outcome shows a linear trend across "ordered groups" (e.g. increasing with dose/grade). More targeted than a plain chi-square — chi-square only asks "is there a difference", the trend test asks "is there a monotone increase or decrease".
① Enter ordered groups
One group per line (in order): score, total, events. The score is the ordered level (e.g. 0,1,2,3 for increasing dose/stage); you can also enter the actual dose value.
How to use & methodology
How does it differ from an ordinary chi-square test?
The k×2 chi-square only judges 'whether the group rates differ overall', ignoring group order; the trend test uses the ordering, specifically testing whether the rate rises or falls monotonically with level — more sensitive and higher-powered for a real dose-response relationship.
How do I set the scores?
If the levels are roughly equally spaced, 0,1,2,3… is common; if there are actual values (e.g. doses 5/10/20 mg), entering the values is more reasonable. The scores represent the ordered spacing between groups, and different choices slightly change the statistic.
What does a non-significant result mean?
It means no significant linear trend was found, but it does not rule out a non-linear (e.g. up-then-down) relationship — the trend test may miss these, so combine it with a plot of the group rates and the chi-square.
How many groups and how large a sample?
At least 3 ordered groups are meaningful; the larger each group, the more stable the test. When events are all 0 or all occur, the test cannot be computed.