Concerning the statistical procedures enumerated by Gentile et al.: another perspective.

Périkel et al. (1974) wrote a short, sharp note about statistics. They said stop using ANOVA F tests on single-case data. Use Bonferroni-corrected t tests instead.

The paper is pure method talk. No kids, no rats, just math advice for behavior analysts.

Bonferroni t keeps the family-wise error low and keeps power high when your data points are correlated. ANOVA loses power in that same spot.

Moeyaert et al. (2014) extends the same worry to meta-analysis. They show how to build design matrices that respect within-case correlation, giving cleaner effect sizes.

Alsop (2004) also warns about bias, but in signal-detection estimates. Both papers say routine fixes can mislead if you ignore how behavioral data behave.

Kessel (2004) offers a concrete fix for noisy records: Wiener filtering. It sharpens transfer-function estimates, echoing the 1974 call for cleaner quantification.

Next time you run a reversal or multiple-baseline graph, skip the ANOVA button. Run separate t tests with a simple Bonferroni correction. You keep power and stay honest about Type I error. It takes one extra line in Excel or R and gives you stronger evidence for your visual analysis.