A tree bootstrap method for estimating uncertainty in respondent-driven samples (RDS). Quantiles are estimated by multilevel resampling in such a way that preserves the dependencies of and accounts for the high variability of the RDS process.
