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

Table 1 Population and mean sample characteristics for each simulated population

Population	Population characteristics				Mean sample characteristics			Sampling correlation^a
Population	Prevalence	Homophily	Odds ratio	Relative risk	Degree	Number of waves	Recruits per seed	Sampling correlation^a
1	10%	1.00	7.59	2.86	44.4	8.4	57.5	0.899
2	10%	1.10	7.65	2.88	43.5	8.3	57.2	0.895
3	10%	1.25	7.22	2.84	44.2	8.4	57.0	0.900
4	10%	1.50	6.93	2.85	43.7	8.3	56.9	0.896
5	30%	1.00	7.47	2.05	43.8	8.1	55.9	0.896
6	30%	1.10	7.56	2.05	43.4	8.1	55.6	0.891
7	30%	1.25	7.47	2.05	44.4	8.2	55.9	0.894
8	30%	1.50	7.59	2.06	44.2	8.2	56.3	0.894
9	50%	1.00	7.47	1.68	43.6	8.2	55.6	0.890
10	50%	1.10	7.55	1.68	43.5	8.1	55.6	0.890
11	50%	1.25	7.50	1.69	44.2	8.2	55.3	0.892
12	50%	1.50	7.51	1.69	44.0	8.2	55.9	0.893

ISSN: 1471-2288