Child eyewitness researchers often bin age: Prevalence of the practice and recommendations for analyzing developmental trends.

Abstract

Objective: Effective practices for eliciting and analyzing children's eyewitness reports rely on accurate conclusions about age differences in how children retain information and respond to memory probes. Binning, which is the practice of categorizing continuous variables into discrete groups, can lower studies' power to detect age differences and, in some situations, produce significant but spurious effects. In this article, we (a) describe a systematic review that estimated the frequency of binning age in child eyewitness studies, (b) analyze real and simulated data to illustrate how binning can distort conclusions about age and covariate effects, and (c) demonstrate best practices for analyzing and reporting age trends. Hypotheses: We expected that researchers would frequently bin age and that we would replicate the negative consequences of binning in the demonstration data sets. Method: For the systematic review, we retrieved 58 articles describing child eyewitness studies and determined whether researchers binned age for one randomly selected analysis per article. We then compared alternative ways of analyzing actual and simulated data sets. Results: Researchers binned age for 64% of the analyses (88% of analyses involving experimental manipulations vs. 35% of the nonexperimental analyses, ϭ .55, p Ͻ .01). A significant age trend in the real data example was nonsignificant when age was treated as categorical, and in the simulated data sets we demonstrate how this practice may lead to detecting a spurious effect. Conclusions: Treating age as a continuous variable maximizes power to detect real differences without inflating the frequency of spurious results, thereby ensuring that policies regarding child eyewitnesses reflect developmental changes in children's needs and abilities. Public Significance StatementResearchers who study children's eyewitness testimony frequently look for developmental trends by comparing children's performance across discrete age groups. Statistical approaches that treat age as a continuous variable are more likely to return accurate conclusions about children's strengths and weaknesses as witnesses, however, and should be the usual way of analyzing developmental trends. An online supplement illustrates this approach through example code and annotated output from SPSS and R.

Altered patterns of brain dynamics linked with body mass index in youth with autism

“…Upon ensuring there were no diagnostic differences in brain‐behavior relationships, we further examined BMI and behavior interactions utilizing a dimensional approach. A dimensional approach was chosen because dichotomizing data decreases power while also inflating Type I error rates (Bainter et al, 2020; MacCallum et al, 2002). Additionally, previous work has shown that although children with ASD are more likely to have impaired social and executive functioning, grouping based on executive function did not reproduce diagnostic groups (Baez et al, 2020; Dajani et al, 2016).…”

Making memories: Examining the gestural misinformation effect in children aged 11-16-years-old with intellectual/developmental difficulties.

“…Following from research demonstrating that the act of 'binning' ages can cause a loss of power for detecting an effect (Bainter et al, 2020)…”

The digital witness: Exploring gestural misinformation in tele-forensic interviews with 5-8-year-old children

“…Analysis included prior work by the authors into the GME in face-to-face interviews using the same methodology and misinformation paradigm to ensure appropriate and reliable comparisons could be made. Age was analysed continuously to avoid the act of 'binning' participants which can create a lack of power when detecting effects (Bainter et al, 2020). was anticipated that even subtle gestures would have a strong misleading in uence but that this would be less than salient gestures due to stronger visual cues engaging the mirror-neuron system (Proverbio & Zani, 2022) and motor system (Ianì & Bucciarelli, 2017) more robustly.…”