Based on published collaborative research co-authored by Dr Arash Ardalan.
The research question
Could routine body measurements help predict segmental body composition in children and adolescents?
Study context
The study used measurements from 476 participants aged 9–18 in southern Iran. Body composition measured by DXA provided a reference against which predictions could be assessed.
Statistical approach
- Fit Bayesian network models using anthropometric predictors, with the bnlearn package in R.
- Develop models on 80% of each study group and evaluate them using the remaining 20%.
- Compare predicted body composition with the reference measurements.
What the analysis provided
The paper reported prediction equations and their performance against DXA measurements, showing how a modelling approach could be assessed within the study sample.
The study concerned a specific population and measurement setting. Use in another population would require further validation.
Practical lessons for your project
- Agree what success means
- Define the outcome, the intended use and the performance measures before choosing a model.
- Plan the evaluation
- Keep model development and evaluation separate, using a validation approach appropriate to the data and study design.
- Make the analysis usable
- An agreed project can include documented code, model checks and a clear account of where the results apply.
Original publication
Mahsa Rahmani; Arash Ardalan; Mostafa Ghaderi-Zefrehei; Marjan Jeddi; Seyed Taghi Heydari; Mohammad Hossein Dabbaghmanesh
Probabilistic prediction of segmental body composition in Iranian children and adolescents
BMC Pediatrics · 22, 524 · 2022