Scalable Bayesian inference for bradley-Terry models with ties: an application to honour based abuse.

Area of Science:

• Public Health
• Criminology
• Social Sciences

Background:

• Honour-based abuse (HBA), including female genital mutilation and forced marriage, requires local data for effective safeguarding.
• Current data limitations hinder accurate identification and prevention of HBA at the community level.

Purpose of the Study:

• To map the local risk of honour-based abuse.
• To address data gaps by employing comparative judgement surveys.
• To develop an efficient computational model for analyzing comparative data.

Main Methods:

• Comparative judgement surveys were conducted, presenting participants with pairs of local areas (wards) to assess relative rates of HBA.
• An efficient Markov Chain Monte Carlo (MCMC) algorithm was designed to fit a statistical model accommodating tied comparisons, reducing participant fatigue.
• The model allowed for flexible prior distributions, enhancing its applicability.

Main Results:

• The study successfully mapped the risk of honour-based abuse at the community level in two UK counties.
• The developed MCMC algorithm efficiently handled tied comparisons, a common issue in comparative judgement studies.
• Collaboration with law enforcement and NGOs facilitated local data collection and risk mapping.

Conclusions:

• Comparative judgement surveys, coupled with an efficient MCMC model, provide a viable method for mapping local HBA risks.
• This approach can support safeguarding professionals in identifying and supporting at-risk populations.
• Addressing local data deficits is crucial for targeted HBA prevention and intervention strategies.