Characteristics of Moderators in Meta-Analyses of Single-Case Experimental Design Studies.
Expect diagnosis, age, and gender to be missing in about one of every four single-case meta-analyses—plan your multilevel model power around that hole.
01Research in Context
What this study did
Moeyaert et al. (2023) read 60 meta-analyses that used single-case designs. They wrote down every moderator the authors tested.
They counted how many papers said each child’s diagnosis, age, or sex. They also noted how often that data was missing.
What they found
Most meta-analyses use only three to four moderators. Diagnosis, age, and gender top the list.
About fifteen to thirty percent of the time this basic info is missing. Planning a multilevel model? Expect holes.
How this fits with other research
Jamshidi et al. (2018) looked at the same pool of meta-analyses but scored their quality. They found many skip bias checks and botch the math. Mariola shows one reason: key facts about kids are simply absent.
Baek et al. (2023) gives a cookbook for multilevel meta-analysis. Their guide pairs with Mariola’s list. Use the moderators Mariola found and plug the missing-data rate into your power code.
Vassos et al. (2023) mined graphs for transition states. Like Mariola, they map hidden patterns that can sway results. Together the two papers form a checklist: watch for slow change and missing moderators before you trust an effect.
Why it matters
When you build a meta-analysis or vet one for practice, you now know which child facts to ask for and how many gaps to expect. Plug the missing-data rate into your multilevel model so your standard errors stay honest.
Want CEUs on This Topic?
The ABA Clubhouse has 60+ free CEUs — live every Wednesday. Ethics, supervision & clinical topics.
Join Free →Open the last SCED meta-analysis you cited and check the moderator table—if age or diagnosis is blank in >20 % of cases, treat the effect size with caution.
02At a glance
03Original abstract
Hierarchical linear modeling (HLM) has been recommended as a meta-analytic technique for the quantitative synthesis of single-case experimental design (SCED) studies. The HLM approach is flexible and can model a variety of different SCED data complexities, such as intervention heterogeneity. A major advantage of using HLM is that participant and-or study characteristics can be incorporated in the model in an attempt to explain intervention heterogeneity. The inclusion of moderators in the context of meta-analysis of SCED studies did not yet receive attention and is in need of methodological research. Prior to extending methodological work validating the hierarchical linear model including moderators at the different levels, an overview of characteristics of moderators typically encountered in the field is needed. This will inform design conditions to be embedded in future methodological studies and ensure that these conditions are realistic and representative for the field of SCED meta-analyses. This study presents the results of systematic review of SCED meta-analyses, with the particular focus on moderator characteristic. The initial search yielded a total of 910 articles and book chapters. After excluding duplicate studies and non peer-reviewed studies, 658 unique peer-reviewed studies were maintained and screened by two independent researchers. Sixty articles met the inclusion criteria and were eligible for data retrieval. The results of the analysis of moderator characteristics retrieved from these 60 meta-analyses are presented. The first part of the results section contains an overview of moderator characteristics per moderator level (within-participant level, participant level, and study level), including the types of moderators, the ratio of the number of moderators relative to the number of units at that level, the measurement scale, and the degree of missing data. The second part of the results section focuses on the metric used to quantify moderator effectiveness and the analysis approach. Based on the results of the systematic review, recommendations are given for conditions to be included in future methodological work.
Behavior modification, 2023 · doi:10.1177/01454455211002111