Snowball samples begin from a core of known elements and are then increased by adding new elements given by members of the original sample. They are so called on the analogy of the increasing size of a snowball when rolled down a snow-covered slope. Such samples are often used where there is no available sampling frame listing all the elements for the population of interest, for example illicit drug-users. Hence snowball samples are not random and not statistically representative of the population under consideration. They are, therefore, not amenable to inferential statistical techniques.