Sampling (statistics) - Nonprobability Sampling Techniques

Nonprobability Sampling What Is Nonprobability Sampling? Nonprobability sampling is a method of selecting sample elements where the probability of selection is unknown or not specified. Unlike probability sampling, where each element has a known chance of being included, nonprobability sampling relies on the researcher's judgment, convenience, or other non-random mechanisms to decide which elements to include. Key idea: The researcher makes choices about who gets included in the sample, rather than letting randomness determine this. Why Use Nonprobability Sampling? Researchers often turn to nonprobability sampling when: The target population is hidden or hard to identify (like undocumented immigrants or people with rare diseases) Random selection is impractical or impossible Budget or time constraints make probability sampling infeasible The research goal is exploratory rather than seeking population estimates However, these convenience benefits come with serious trade-offs that you need to understand. Critical Limitations of Nonprobability Sampling Unable to Estimate Sampling Error The most fundamental limitation of nonprobability sampling is that we cannot calculate sampling error or confidence intervals. Here's why: Sampling error depends on knowing the probability that each element was selected. For example, if element A had a 1-in-50 chance of selection and element B had a 1-in-100 chance, those different odds affect how representative the sample is likely to be. With nonprobability sampling, we don't know these odds—sometimes we don't even know if certain elements had any chance of being selected at all. What this means for interpretation: You cannot say "our results are accurate within ±5 percentage points with 95% confidence" when using nonprobability sampling. Any statistical confidence intervals reported from nonprobability samples are misleading. Exclusion Bias Exclusion bias occurs when some elements in the population have zero probability of selection—they are "out of coverage." These excluded elements may be systematically different from those who can be reached, which limits how well you can generalize your findings. Example: If you conduct a convenience sample by standing outside a grocery store on a weekday morning, you systematically exclude working people who can't shop at that time, shift workers, people with mobility limitations, and those who shop online. Your sample will be biased toward people available during that specific time. Exclusion bias is particularly problematic because you may not even know what you're missing. How Nonresponse Converts Probability Designs into Nonprobability Designs This is a subtle but important point: a probability design can become a nonprobability design if nonresponse is severe and the nonrespondents differ systematically from respondents. Imagine you start with a proper random sample (probability design). You send surveys to 1,000 randomly selected people, but only 200 respond. If those 200 respondents are fundamentally different from the 800 nonrespondents—say, they tend to be more educated or more engaged—then your effective sample is no longer representative. The nonrespondents have effectively been excluded, converting your design into a nonprobability situation. Key insight: Nonresponse is a hidden source of exclusion bias. When you can't assess how nonrespondents differ from respondents, your probability design loses its validity. Common Nonprobability Sampling Techniques Convenience Sampling Convenience sampling selects elements that are easiest or most readily available to the researcher. Examples: A university researcher studies students in their own classes A mall intercept survey asks shoppers who happen to be present An online survey posted to social media reaches whoever clicks the link Strengths: Quick and inexpensive Weaknesses: Highly susceptible to selection bias; results may not generalize beyond the immediate accessible population Quota Sampling Quota sampling divides the population into groups (called strata) based on known characteristics, then fills each group with a predetermined number of elements. The researcher has discretion in who within each group gets selected. Example: A researcher wants a sample that matches the population's age distribution. They decide to include 25% ages 18-30, 35% ages 31-50, and 40% ages 51+. However, within each age group, they use convenience to select the specific people. Key distinction: While quota sampling resembles stratified random sampling, it differs because selection within quotas is not random. This introduces bias. Judgmental (Purposive) Sampling Judgmental sampling relies on an expert's knowledge and judgment to select participants expected to provide especially valuable or representative information. Examples: A hospital researcher studying a new treatment selects patients they believe represent different disease severity levels A food scientist testing a new recipe recruits experienced home cooks and professional chefs A marketing team studying luxury car preferences recruits people with high incomes Strength: Can efficiently target information-rich cases Weakness: Highly dependent on the expert's judgment; easy for unconscious bias to influence selection Snowball Sampling Snowball sampling starts with a small initial group of respondents, who then recruit additional participants from among their acquaintances. The sample "grows" like a rolling snowball. Why use it: Snowball sampling is invaluable for hidden or hard-to-reach populations—those without a sampling frame or public list. Examples: Studying undocumented immigrants: initial contacts refer friends and family Researching people with a stigmatized condition: patients connect the researcher with others willing to participate Studying homeless populations: street outreach workers introduce the researcher to other individuals Important limitation: People tend to refer others similar to themselves, so snowball samples often become less diverse the further the chain extends. The later respondents may not represent the full variation in the hidden population. Advanced Nonprobability Techniques Theoretical Sampling Theoretical sampling selects cases strategically based on emerging findings to deepen understanding of a phenomenon. This approach is common in qualitative research. How it works: Begin with a broad initial sample to understand general patterns Analyze the data and identify gaps or interesting variations Deliberately add new cases that represent extreme values, contrasts, or specific conditions to illuminate the phenomenon more fully Example: A researcher studying why employees leave certain jobs starts by interviewing 10 people with average tenure at a company. They notice that people who left extremely quickly (within 3 months) had different experiences than those who stayed 2-3 years. They then deliberately seek out more people in both extreme groups to understand what creates these differences. Key purpose: Not to estimate population parameters, but to develop rich understanding and theory. Haphazard Sampling Haphazard sampling uses human judgment to mimic randomness when formal random selection tools are unavailable. Example: A researcher studying homeless individuals in a city might walk different streets and times of day, trying to encounter different types of homeless people, rather than standing in one location. Strengths: Practical when formal lists don't exist Weaknesses: Easily biased by researcher preferences; researcher may unconsciously seek out cooperative or articulate participants; cannot truly mimic true randomness <extrainfo> Less Common or Specialized Nonprobability Techniques Other nonprobability methods you might encounter include: Network sampling: Similar to snowball sampling but more systematized for studying social networks Respondent-driven sampling (RDS): A sophisticated version of snowball sampling that attempts to weight results to compensate for the biased recruitment chains These are more specialized and less likely to be primary exam content, but they represent important variations in how researchers adapt nonprobability methods for specific research challenges. </extrainfo> Summary: When to Use Nonprobability Sampling Nonprobability sampling is appropriate when: You're exploring a new topic (not estimating population parameters) The population is hidden or difficult to enumerate You need rich, detailed information from specific individuals Practical constraints make probability sampling infeasible Nonprobability sampling is not appropriate when: You need to estimate population characteristics with known precision You must generalize findings to a clearly defined population You need to report confidence intervals or margins of error The fundamental trade-off: Nonprobability sampling sacrifices external validity and statistical precision for feasibility and depth of understanding.