multiple baseline designs

[ASKurz]

Please leave suggestions for single-case or small-n studies using a multiple baseline design. I'm primarily thinking about multiple baselines across persons, but these could also be multiple baselines across settings or behaviors. These could also be redundant with suggestions for ABAB designs, and so on.

[ASKurz]

Consider Himle et al (2013; https://doi.org/10.1901/jaba.2004.37-1), Teaching safety skills to children to prevent gun play. Here's the abstract:

Research has shown that children often engage in gun play when they find a firearm and that this behavior is often involved in unintentional firearm injuries. Previous research has shown existing programs to be ineffective for teaching children safety skills to reduce gun play. This study examined the effectiveness of a behavioral skills training (BST) program supplemented with in situ training for teaching children safety skills to use when they find a gun (i.e., don't touch, leave the area, tell an adult). Eight 4- to 5-year-old children were trained and assessed in a naturalistic setting and in a generalized setting in a multiple baseline across subjects design. Results showed that 3 of the children performed the skills after receiving BST, whereas 5 of the children required supplemental in situ training. All children in the study learned to perform the skills when assessed in a naturalistic setting and when assessed in a generalization setting. Performance was maintained at 2- to 8-week follow-up assessments.

This is an n = 8 multiple baseline AB design where some of the children had subsequent in-situ training after the B phase. The primary outcome is a 4-point ordinal scale, which would give a nice example for a softmax analysis. This article was used in Kate's class (file path: /BA Seminar/7. Increasing Community Safety/Himle et al., 2004 JABA.pdf).

[ASKurz]

Consider Donaldson et al (2015; https://doi.org/10.1002/jaba.229), Immediate and distal effects of the good behavior game. Here's the abstract:

The Good Behavior Game (GBG) has been demonstrated to reduce disruptive student behavior during implementation. The effects of playing the GBG on disruption immediately before and after the GBG are unknown. The current study evaluated the effects of the GBG on disruption in 5 kindergarten classes immediately before, during, and after GBG implementation. The GBG reduced disruption during implementation but did not affect rates of disruption during activity periods that preceded or followed the GBG.

Here n = 5 kindergarten classes participated in a "two-component multiple-schedule arrangement with reversals across components to examine the effects of the GBG and standard teacher contingencies (STC) on student disruptions" (p. 686). The two outcomes are counts per minute and the analysis would be a nice challenge. This article was used in Kate's class (file path: /BA Seminar/3. Behavior Analysis & Schools/3. Lagniappe/Donaldson et al., 2015.pdf).

[ASKurz]

Consider Li et al (2024; https://doi.org/10.3758/s13428-024-02359-7), Multilevel modeling in single-case studies with zero-inflated and overdispersed count data. Here's the abstract:

Count outcomes are frequently encountered in single-case experimental designs (SCEDs). Generalized linear mixed models (GLMMs) have shown promise in handling overdispersed count data. However, the presence of excessive zeros in the baseline phase of SCEDs introduces a more complex issue known as zero-inflation, often overlooked by researchers. This study aimed to deal with zero-inflated and overdispersed count data within a multiple-baseline design (MBD) in single-case studies. It examined the performance of various GLMMs (Poisson, negative binomial [NB], zero-inflated Poisson [ZIP], and zero-inflated negative binomial [ZINB] models) in estimating treatment effects and generating inferential statistics. Additionally, a real example was used to demonstrate the analysis of zero-inflated and overdispersed count data. The simulation results indicated that the ZINB model provided accurate estimates for treatment effects, while the other three models yielded biased estimates. The inferential statistics obtained from the ZINB model were reliable when the baseline rate was low. However, when the data were overdispersed but not zero-inflated, both the ZINB and ZIP models exhibited poor performance in accurately estimating treatment effects. These findings contribute to our understanding of using GLMMs to handle zero-inflated and overdispersed count data in SCEDs. The implications, limitations, and future research directions are also discussed.

Their empirical example data are from an $N = 5$ multiple-baseline AB design. Most participants have very few occasions per phase, making this a good example to discuss intercepts-only versus intercepts-and-slopes models, and the methodological issue of how many occasions one should have per phase. The authors made their data open on the OSF at https://osf.io/e4tp5/. In addition to contrasting Poisson, NB, ZIPoisson and ZINB models, this is a great opportunity to fit all them all as full distributional models.