The statistical analysis of surveys using representative samples usually starts with computing mean scores on key variables of interest. Sample means can be of particular interest when exploring changes over time and differences across national samples. But full sample means can be misleading when subpopulations differ systematically, such as when Latvian and Russian speaking Latvians differ so dramatically in their views of Russia and China. Secondary survey analysis can also explore demographic and other (e.g. ideological) drivers of key dependent measures (such as China attitudes)—as well as potential mediators and moderators of such effects. Unlike deductive theory-driven research, this approach requires an inductive commitment to the data—and keeping an eye on policy significance and extant arguments and silences in the literature.