Unweighted regression models perform better than weighted regression techniques for respondent-driven sampling data: results from a simulation study

Table 3 Outcome prevalence estimates using various estimators across populations

Homophily:	Outcome prevalence 10%				Outcome prevalence 30%				Outcome prevalence 50%
Homophily:	1.00	1.10	1.25	1.50	1.00	1.10	1.25	1.50	1.00	1.10	1.25	1.50
Mean outcome prevalence
naïve	0.09	0.09	0.09	0.09	0.27	0.27	0.27	0.27	0.47	0.47	0.47	0.46
RDS-I	0.08	0.08	0.08	0.08	0.27	0.26	0.26	0.26	0.47	0.47	0.46	0.46
RDS-II	0.08	0.08	0.08	0.08	0.27	0.26	0.26	0.26	0.47	0.47	0.46	0.46
surveylogistic models
unweighted	0.09	0.09	0.09	0.09	0.27	0.27	0.27	0.27	0.47	0.47	0.47	0.46
weighted (RDS-II)	0.08	0.08	0.08	0.08	0.27	0.26	0.26	0.26	0.47	0.46	0.46	0.45
Mean SD of outcome prevalence
naive	0.01	0.01	0.01	0.02	0.02	0.02	0.02	0.03	0.02	0.02	0.03	0.03
RDS-I	0.02	0.02	0.02	0.03	0.04	0.04	0.04	0.04	0.04	0.05	0.05	0.05
RDS-II	0.02	0.02	0.02	0.03	0.04	0.04	0.04	0.05	0.04	0.05	0.05	0.05
surveylogistic models
unweighted	0.01	0.01	0.01	0.02	0.02	0.02	0.02	0.03	0.02	0.02	0.03	0.03
weighted (RDS-II)	0.02	0.02	0.02	0.03	0.04	0.04	0.04	0.04	0.04	0.05	0.05	0.05
Estimator coverage rates
naive	0.845	0.827	0.802	0.708	0.646	0.740	0.620	0.642	0.742	0.687	0.634	0.551
RDS-I	0.545	0.554	0.548	0.578	0.572	0.512	0.524	0.501	0.627	0.610	0.569	0.511
RDS-II	0.772	0.776	0.766	0.749	0.799	0.761	0.744	0.723	0.839	0.831	0.791	0.741
surveylogistic models
unweighted	0.916	0.900	0.875	0.784	0.657	0.745	0.611	0.645	0.747	0.684	0.644	0.544
weighted (RDS-II)	0.828	0.819	0.799	0.769	0.825	0.779	0.778	0.753	0.862	0.835	0.819	0.756

ISSN: 1471-2288