At the beginning of the 20th century, there was an active debate about random selection of units versus purposive selection of groups of units for survey samples. Neyman's (1934) paper tilted the balance strongly towards varieties of probability sampling combined with design-based inference, and...