SNA's origins in sociometry mean that most of its datasets and templates are (or are presented as) complete, whole network data where all respondents on a roster have indicated the existence, or not, (and/or strength) of their ties to each and every other person on the roster. Data collection is seldom this complete and collecting sociometric data across even moderate sized study populations makes the use of a full roster impractical. This paper considers key issues for research design when we cannot expect to collect complete sociometric data. It examines major SNA studies which work with partial
sociometric data and their justifications of research design and data analysis issues. It considers the use of proxy data collection in Ron Burt's studies and other SNA research. Finally it looks at emerging techniques of data imputation and their justifications and potential. The paper argues that de facto use of partial sociometric research designs is such that a more systematic recognition and specification of it is now called for.