This document proposes a new statistical framework for modeling, analyzing, and evaluating anonymity in sensor networks. The framework introduces the notion of "interval indistinguishability" to quantitatively measure anonymity. It maps the source anonymity problem to the statistical problem of binary hypothesis testing with nuisance parameters. Existing solutions are analyzed using this framework, showing how anonymity can be improved by finding an appropriate data transformation to remove nuisance information. Mapping the problem to binary hypothesis testing opens opportunities to apply coding theory to anonymous sensor networks.