Some methodological issues in the study of sexual networks: from model to data to model.

ABSTRACT

BACKGROUND: Mixing between sexual activity classes is an important determinant of sexually transmitted disease transmission. However, attempts to estimate sexual mixing patterns in the field remain limited partly because of practical and methodological difficulties. GOAL To evaluate and identify appropriate sampling schemes to estimate the mixing pattern between sexual activity classes from large population networks with one or more components. STUDY DESIGN: The study is based on simulations of large population networks with various structural characteristics. A variety of snowball sampling schemes are applied to these networks and are evaluated by the quality of the mixing matrix estimates that they produce. RESULTS AND CONCLUSIONS: Unbiased estimation of mixing patterns (global assortativity, within-group mixing of the lowest activity classes, within-group mixing of the highest activity classes) from large population networks is possible with a snowball sampling design in which the initial sample of index cases is drawn from the general population, all partners of the index case are recruited, and only one generation of partners are traced (one cycle). Simulation techniques proved useful in addressing complex methodological issues in situations where analytic results are difficult to obtain.