ABAB designs

[ASKurz]

Please leave suggestions for single-case or small-n studies using an ABAB design. These could also include variants, such as with repeated reversals (e.g., ABABAB), the BABA design, and multiple-treatment reversal designs (e.g., ABAC).

[ASKurz]

Consider Smith & Ward (2006; https://doi.org/10.1901/jaba.2006.5-06), Behavioral interventions to improve performance in collegiate football. Here's the abstract:

Using a multitreatment withdrawal design, this study evaluated the differential effects of publicly posted plus verbal feedback, goal setting plus verbal feedback, and publicly posted feedback, verbal feedback, and goal setting together on the performance of 3 collegiate football players in practice scrimmages. Also assessed was whether the changes in practice behavior generalized to games. The dependent variables were performances on three wide receiver skills. The results show that public posting with verbal feedback, goal setting, and public posting with verbal feedback and goal setting were effective in improving player performance to a 90% criterion level during practice, and these changes generalized to game performance.

This is an n = 3 ABACABC multitreatment withdrawal design, where the last phase is BC, which combined elements of B and C. There are 3 criterion variables, all of which are a percent correct out of 10. However, putting the cap at 10 was arbitrary and could lead into a discussion of different n across participants when using the binomial likelihood. Also, this article was used in Kate's class (file path: /BA Seminar/10. Sports and Leisure/Smith & Ward, 2006.pdf).

[ASKurz]

Consider Greene et al (1981; https://doi.org/10.1901/jaba.1981.14-177), An analysis and reduction of disruptive behavior on school buses. Here's the abstract:

Thousands of children are injured or killed each year in school bus accidents. A significant number of these tragic incidents is precipitated by disruptive child behavior that distracts the drivers from their difficult task. Two experiments were conducted which addressed this problem. For both experiments an automated sound recording device (referred to as a Noise Guard) selectively responsive to frequencies above 500 Hz (i.e., unresponsive to bus drone) recorded both the duration and frequency of noise outbursts above a tolerable threshold. Additionally, an observer made in situ measurements of other disruptions including roughhousing and getting-out-of-seat. In the first experiment, following baseline measurements of these behaviors, middle-school students received feedback for noise outbursts. That is, when “Noise Guard'” was activated, it in turn operated one of several lights on a panel visible to all passengers. Each day students were allowed to listen to high-appeal taped music while riding the bus and to participate in a raffle for prizes, provided the number of outbursts on the preceding day remained below a specified criterion indicated on the light panel. This intervention resulted in drastic reductions of noise outbursts with a concomitant reduction in other disruptive behaviors. Comparable results were obtained in the second experiment which eliminated the raffle from the intervention.

This is an n = 2 multiple baseline ABABC design with multiple outcomes. For simplicity, you might just focus on the two outcomes for "Bus #70" presented in figure 1. They would give a nice example of single-case models using the count regression and response-duration regression (e.g., gamma, lognormal). This article was used in Kate's class (file path: /BA Seminar/3. Behavior Analysis & Schools/Greene, Bailey, & Barber, 1981.pdf).

[ASKurz]

Consider Lambert et al (2006; https://doi.org/10.1177/10983007060080020701), Effects of response cards on disruptive behavior and academic responding during math lessons by fourth-grade urban students. Here's the abstract:

The authors evaluated the effects of response cards on the disruptive behavior and academic responding of students in two urban fourth-grade classrooms. Two conditions, single-student responding and write-on response cards, were alternated in an ABAB design. During single-student responding, the teacher called on one student who had raised his or her hand to answer the question. During the response-card condition, each student was provided with a white laminated board on which he or she could write a response to every question posed by the teacher. Nine students were targeted for data collection because of their history of disciplinary issues in school and frequent disruptive behavior in the classroom. Data revealed substantial reductions in disruptive behavior and increases in academic responding during the response card condition compared to single-student responding. The findings are discussed in terms of the beneficial effects of direct, high-response strategies for urban, low-achieving learners.

This is an n = 9 ABAB reversal design with a bounded count outcome. Students are nested in two classes, 4 in one and 5 in the other, which makes for a nice complication. There's an additional complication that the initial A1 phase has very minor between-student differences, which affords a discussion of whether to keep the full random structure. There's also the issue that many other studies have used these data to compare analytic strategies (e.g., Manolov & Ferron 2020; Michiels & Onghena, 2019; Moeyaert et al, 2014; Peng & Chen, 2015; Shadish et al, 2013; Shadish et al, 2014). When writing this up, the walk-out in Shadish et al (2013) is particularly relevant and should probably be explicitly discussed. As the prior studies generally didn't emphasize it, you should also highlight the issue of nesting within classrooms and explicitly build that into the analysis, with references to Gelman et al's discussion of group/cluster-randomized experiments (2020, Section 18.5).

[ASKurz]

Consider Keyes et al (2019; https://doi.org/10.1016/j.jbtep.2019.101523), Time intensive cognitive behavioural therapy for a specific phobia of vomiting: A single case experimental design. Here's the abstract:

Background and Objectives Specific Phobia of Vomiting (SPOV) may be difficult to treat, and to date only one RCT has been published. Specific Phobias can be successfully treated using time intensive treatment formats. Imagery rescripting (ImRs) has also been shown to be a successful component for treating various anxiety disorders. The present study aims to evaluate the effectiveness of time intensive CBT and ImRs at reducing symptoms in SPOV.

Methods A multiple baseline ABCADE single case experimental design (SCED) was used to monitor symptoms across time in eight participants. The Specific Phobia of Vomiting Inventory (SPOVI) was administered to measure SPOV symptoms at each clinical session. Visual analysis of graphed data was used to compare data across treatment phases. Reliable and clinically significant change was also calculated on the SPOVI at 6-month follow up.

Results Findings suggest that time intensive CBT is associated with improvements across symptoms of SPOV. There was limited support for ImRs in the current study; however, informal discussion of aversive memories of vomiting, and formulation of the problem may be helpful in its own right in treatment of SPOV. Seven out of eight participants (87.5%) achieved reliable improvement and five (62.5%) achieved clinically significant change on the SPOVI at 6-month follow up. Time intensive treatment was associated with high client satisfaction ratings.

Limitations Limitations include lack of measurement of therapist adherence to protocol, and a short period of daily symptom monitoring in between ImRs and intensive treatment phases.

Conclusions Time intensive CBT including an ImRs component may be an effective and acceptable treatment in reducing SPOV symptomatology. Further research using larger sample sizes in a RCT is needed.

This is an n = 8 multiple baseline ABCADE design with two primary outcomes. Both were measured with a visual analogue scale ranging from 1 to 10, which would make for a nice example for the zero-one-inflated beta distribution. You might even model both DVs with a bivariate ZIOB model.

[ASKurz]

Consider Thorne & Kamps (2008; https://doi.org/10.1007/BF03391723), The effects of a group contingency intervention on academic engagement and problem behavior of at-risk students. Here's the abstract:

A successful learning environment can be characterized by actively engaged students displaying appropriate student behavior. We implemented a group contingency intervention as a novel component to a school-wide behavior management system to decrease the frequency of inappropriate behaviors and, conversely, increase the academic engagement of students in four elementary school classrooms. Twelve students with behavioral risks served as target students to monitor effects. A reversal design was implemented to evaluate behaviors across experimental conditions. Results indicated that the frequency of inappropriate behaviors decreased and academic engaged time increased for all 12 participants. These results suggested that the group contingency was an effective class-wide intervention. Implications for research and practice are discussed.

This is an n = 12 ABAB reversal design with two primary outcomes, a % on-task behavior (perhaps use ZIOB) and a rate (# behaviors within a 15-minute time frame, which could be transformed back to a count and modeled as Poisson or NB). Students are clustered into 4 classroom settings, making for a 3-level design. The data are also discussed in Swan & Pustejovsky (2018; https://doi.org/10.1080/00273171.2018.1466681).