Researchers often need to measure new variables, or validate existing measurements. Doing this for a whole cohort or database can be prohibitively expensive, so techniques are needed for choosing good subsamples and analyzing them efficiently. This module will give both a conceptual and practical introduction to planning and analyzing modern two-phase study designs. We will cover efficient weighted estimation using the whole cohort, and the optimal design and allocation of subsamples. Exercises will focus on practical aspects of implementing methods using the R survey package.