Statistical inference in behavior analysis: Discussant's remarks.

Moerk (1999) is not an experiment. It is a discussant's talk that warns behavior analysts about five traps in statistical inference.

The paper says we often over-trust p-values, ignore effect size, and forget that group averages can hide individual curves.

There is no new data. The piece is a caution flag: if you use t-tests, ANOVAs, or confidence intervals, check that the numbers match what you can see in each single subject's chart.

Frankot et al. (2024) extends L's worry. They show that big-data tools like mixed models and machine learning can keep each person's curve while still giving group-level power.

Fraley (1998) came one year earlier and gave concrete options—matching-to-sample plus feedback—to study judgment. Moerk (1999) then asked how far we should push the stats on those options.

McGeown et al. (2013) and Furrebøe et al. (2017) promote demand-curve metrics from behavioral economics. Moerk (1999) would count these as statistical inference and remind us to test them one subject at a time.

Before you run a t-test on your next treatment study, graph each client's data. If the curves do not all move the same way, skip the group mean and stick with single-case logic.