Snowball samples
• Method: Start with one person with a certain set of characteristics. Use that person to find other people with that characteristic.
• Example: Rare populations (e.g., persons who are HIV+) • Problems:
• No probability mechanism for initial selection, so you don’t know whether your group adequately reflects the whole population of interest
• How you select your first set of respondents affects the entire sample.
• Snowball sampling has reemerged under the guise of “respondent driven sampling.” This is just snowball sampling with fancy formulas, but still has no guarantee of accurately reflecting this population.