Time-series analysis in operant research.

Jones et al. (1977) wrote a how-to guide. They showed ways to test single-case graphs for real change.

The paper gives formulas for level shifts, trends, and autocorrelation. No new data were collected.

The authors found that eyes can fool you. Visual peaks may look big but fail a math check.

Time-series tests give yes-or-no answers. They help decide if the line really moved after you started treatment.

Manolov et al. (2017) extends this work. They turned the 1977 formulas into free R code you can run today.

Wolfe et al. (2019) is a successor. They built step-by-step visual rules so teams can score graphs with numbers instead of guesswork.

Irvin et al. (1998) seems to clash but doesn’t. They showed scatter plots miss patterns that control charts catch. Both papers push the same idea: add stats to your eyes.

Next time you review a client’s A-B graph, run a quick level-shift test before you write the report. Free tools now exist, so you can back up visual calls with numbers in five minutes. This protects you from false positives and helps insurance reviewers trust your data.