A Correlated Network Scale-up Model: Finding the Connection Between Subpopulations.

Area of Science:

• Social Network Analysis
• Epidemiology
• Statistical Modeling

Background:

• Aggregated relational data (ARD) is valuable for network analysis with incomplete data.
• ARD bypasses traditional survey limitations, aiding studies on marginalized groups like female sex workers.
• Previous methods lacked robust analysis for complex network structures and bias reduction.

Purpose of the Study:

• To introduce a novel statistical model for analyzing Aggregated Relational Data (ARD).
• To estimate the size of HIV-related groups in Ukraine using ARD.
• To understand network characteristics influencing social connections and group affiliations.

Main Methods:

• Collected ARD from 10,866 respondents in Ukraine (Dec 2008-Feb 2009).
• Developed a new ARD model incorporating respondent/group covariates and a correlated bias term.
• Introduced a new scaling procedure using correlation structure to minimize bias.

Main Results:

• The proposed model provides improved estimates for the size of at-risk populations for HIV.
• The model enhances understanding of factors influencing social ties and inter-group relationships.
• Bias reduction techniques were successfully applied to ARD analysis.

Conclusions:

• The new ARD model improves the accuracy of network size estimates for HIV prevention.
• Understanding network features aids in recruiting marginalized populations into health programs.
• The model and existing NSUM models are available in the networkscaleup R package.