Hi Folks
I have been enjoying reading the excellent questions and replies about ordinal models in brms.
I am conducting a meta-analysis of studies on adverse childhood experience (ACES). The results of studies using the ACES questionnaire almost exclusively report their data as the counts of study participants who report having 0, 1, 2, 3, or 4+ ACES.
I am interested in the first instance in estimating the average proportion for each category of ACES, as well as tau values representing the average difference in each category between studies.
Meta-analysis of this kind have usually taken the approach proposed by Barendregt et al (2013): Meta-analysis of prevalence - PubMed
The MetaXL extension for Excel handles the approach, but it won’t compute lower confidence intervals for my data…which makes me nervous about using the approach. Also, the approach has been shown to perform less optimally than a Bayesian approach, as shown by Avci (2021): https://www.tandfonline.com/doi/abs/10.1080/03610918.2021.1887229?journalCode=lssp20#:~:text=Multiple%20category%20(multicategory)%20prevalence%20represents,mild%2C%20moderate%2C%20and%20severe.
Avci was unwilling to share their code, so, to me, the paper is a bit of a black box.
I assume that a hierarchical ordinal model, with either the sequential or adjacent category approach, seems like a plausible route forward for me in brms. However, I have not yet found a brms hierarchical model tutorial, paper, or forum post that might help guide me through the meta-analysis.
Does anyone have any suggestions about how to proceed, or suggested resources that they think would fit well with the data I am working with and question I am interested in? Moderation analysis is also of interest, but I want to get an intercepts-only model going first.
Thanks
Ross