As there was very little missing data amongst the factors modelled (≤ 3.1%), complete case analyses were performed. RDS weights were not included, as these may be unwarranted in regression modelling [65]. Robust standard errors were used to account for clustering by each recruiter within ...
Individuals recruit from their personal network at random and recruitment ties are reciprocal, such that one’s likelihood of being recruited to the study is proportional to degree; 4. Recruitment patterns depend only on the recruiter and not on the recruiter’s recruiter; 5. Cross-group ...