The set of tools for thinking hard about sampling and measurement-level aspects of scientific studies is among the earliest areas of statistics to be worked out. However, it arguably is not among the sexier topic areas. In network analysis, perhaps partly as a result of this fact, it is not infrequent that we find ourselves inclined to quickly move past the initial sampling and measurement aspects of the problem (i.e, so-called low-level tasks) in our rush to do things like prediction, detection, and the like (i.e., so-called high-level tasks). In this talk I will present results from work in our group on three problems, one each from social, communication, and biological network analysis. In each case, the goal will be to demonstrate how high-level tasks in network analysis can be fundamentally affected by how faithfully we account for the low-level sampling and/or measurement aspects of the science at hand.