Using the dual-criteria methods to supplement visual inspection: An analysis of nonsimulated data

Lanovaz and his team looked at real single-case graphs from published studies. They wanted to know how many data points you need in each phase before the dual-criteria method gives a safe answer.

They ran the dual-criteria rule on 250 real A-B graphs. They counted how often the rule cried “effect” when the visual picture was really flat.

If you have at least 3 points in baseline and 5 in treatment, the false-alarm rate stays under a large share. With fewer points, the rule starts yelling “effect” too often.

In short: 3-in-A and 5-in-B is the safety line. Cross it before you trust the dual-criteria dots.

Manolov et al. (2022) built a free Brinley-plot tool that uses the same 3-and-5 rule to judge replication across kids. Their tool turns the Lanovaz numbers into a quick picture check.

Manolov (2026) went further and put the rule into one-click web apps. You now upload your file and the site auto-counts your phase points and flags short phases.

Jacobs (2019) pushes randomization tests instead of dual-criteria. Both want to guard visual analysis, but Jacobs uses shuffle stats while Lanovaz uses simple point counts. Pick one; just don’t rely on eyes alone.

Next time you run an A-B design, count your points before you call it a win. If baseline has only two dots or treatment has four, keep collecting. The 3-and-5 rule is fast, free, and keeps your false positives low. Your supervisor and future reviewers will thank you.