Rank‐permutation tests for behavior analysis, and a test for trend allowing unequal data numbers for each subject

Elliffe et al. (2019) built a new trend test for single-case data. The test uses rank-permutation logic. It works even when each client gives a different number of data points.

The authors give free code so you can run the test in R. They also show how it beats older parametric tests when data are skewed or measured on a small scale.

The paper is a how-to guide, not an outcome study. It shows the test keeps the false-positive rate low while still catching real trends.

Killeen (1978) gave rules for knowing when baseline is stable. Elliffe’s test is the next step: once the line is flat, use the rank test to see if the next phase really bends.

Vos et al. (2013) compared two ways to measure response-stimulus links. Elliffe does the same job for trend: it lines up several options and tells you which one to pick.

Baek et al. (2023) teach multilevel modeling for meta-analysis. Their guide and Elliffe’s test are sister tools: one helps pool many small graphs, the other keeps each single graph honest.

If you run single-case sessions and the numbers look messy, skip the t-test. Plug the data into the free rank-permutation test instead. It keeps your error rate clean without forcing you to drop clients who missed a few sessions.