Meta‐analyses and effect sizes in applied behavior analysis: A review and discussion

Dowdy and colleagues walked readers through the growing pile of ABA meta-analyses that use single-case data. They did not run a new meta-analysis. Instead, they explained why each common effect-size index exists and when it can mislead you.

The paper is a roadmap. It compares Tau, Tau-U, PND, PEM, NAP, BC-SMD, and multilevel options. It flags assumptions you must check before you pick one.

No single index wins every race. Each has trade-offs between power, bias, and ease of use. The authors warn that many published meta-analyses ignore these trade-offs, so readers may over-trust small or shaky effects.

Wang et al. (2013) and Hutchins et al. (2020) show the upside: when you pick the right index, social-skills meta-analyses report strong, clear effects. Dowdy et al. (2021) says the clarity can vanish if you switch to a different index or skip assumption checks.

Costello et al. (2022) extends the warning. They ran head-to-head tests and found visual inspection alone often misses real effects that Tau z and RD catch. Their data back Dowdy’s advice to pair effect sizes with small-stat tests, not graphs alone.

Manolov et al. (2025) pushes further, giving step-by-step rules for choosing multilevel terms. Together, the four papers form a timeline: Dowdy sounds the alarm, Costello proves the risk, and Manolov hands you the fix.

If you read or write single-case meta-analyses, Dowdy et al. (2021) is your checklist. Before you believe a large Tau-U, ask if baseline trend was removed. Before you pool PEM across studies, check if baseline variance differs. Share this paper with your next literature-review team so everyone picks the same index up front. One concrete habit: add a short table that lists the index, its assumption test, and the result. Reviewers love it and you safeguard your conclusions.