Molecular epidemiology is increasingly used to investigate patterns of HIV transmission. To do so, many analyses consider investigating properties of a sexual or transmission network. The use of sampled data to estimate such properties is a common practice; however, in the presence of missing data, even missing completely at random, networks based on sampled data do not represent their population counterparts. As a result, inferences on sampled networks become unreliable. To address this challenge, we propose statistical approaches to accommodating missing data in the analysis of sampled networks.