Sampling designs are the backbone of survey research, shaping how we gather data from populations. From simple random sampling to complex multistage methods, each approach has its strengths and weaknesses. Understanding these designs is crucial for collecting representative data and making valid inferences.

Probability sampling ensures every unit has a known chance of selection, while non-probability methods rely on researcher judgment or convenience. The choice between these approaches depends on research goals, resources, and population characteristics. Mastering sampling designs empowers researchers to make informed decisions and conduct robust studies.

Probability Sampling

Simple and Stratified Random Sampling

• Simple random sampling selects units from the population with equal probability
  • Involves assigning numbers to each unit and using a random number generator
  • Ensures every member of the population has an equal chance of selection
  • Requires a complete sampling frame of the entire population
  • Works well for homogeneous populations (student body of a small college)
• Stratified random sampling divides the population into subgroups before sampling
  • Subgroups (strata) are mutually exclusive and collectively exhaustive
  • Samples are drawn independently from each stratum
  • Improves precision for heterogeneous populations (employees in a large corporation)
  • Allows for different sampling methods or sample sizes within each stratum
  • Formula for sample size allocation: $n_h = n \times \frac{N_h}{N}$
    • Where $n_h$ is the sample size for stratum h, $n$ is the total sample size, $N_h$ is the population size for stratum h, and $N$ is the total population size

Cluster and Systematic Sampling

• Cluster sampling selects groups of units rather than individual units
  • Divides the population into clusters, typically based on geographic areas
  • Randomly selects a subset of clusters
  • All units within selected clusters are included in the sample
  • Reduces travel costs for in-person surveys (neighborhoods in a city)
  • Can lead to higher sampling error if clusters are not representative
• Systematic sampling selects units at fixed intervals after a random start
  • Calculates sampling interval k = N/n, where N is population size and n is sample size
  • Randomly selects a starting point between 1 and k
  • Selects every kth unit thereafter
  • Easy to implement and can provide good coverage of the population
  • Works well for ordered lists (selecting every 10th customer from a database)

Multistage Sampling

• Multistage sampling combines multiple sampling methods in successive stages
  • Often used for large-scale surveys covering wide geographic areas
  • First stage typically involves cluster sampling (selecting counties)
  • Subsequent stages may use other methods (simple random sampling within selected counties)
  • Allows for efficient sampling of geographically dispersed populations
  • Can be adjusted at each stage to balance cost and precision
  • Commonly used in national surveys (U.S. Census Bureau's American Community Survey)
• Provides flexibility in sample design and allocation of resources
  • Can concentrate resources on a subset of primary sampling units
  • Allows for different sampling fractions at different stages
• Requires careful planning and weighting to ensure representativeness

Non-Probability Sampling

Convenience and Purposive Sampling

• Convenience sampling selects readily available units
  • Based on ease of access rather than random selection
  • Quick and inexpensive method for gathering data
  • Often used in pilot studies or exploratory research
  • Highly susceptible to selection bias and limits generalizability
  • Commonly employed in street interviews or online surveys
• Purposive sampling selects units based on the researcher's judgment
  • Also known as judgmental sampling
  • Aims to include specific characteristics or expertise in the sample
  • Used when studying rare populations or specific phenomena
  • Allows for in-depth exploration of particular cases or perspectives
  • Frequently utilized in qualitative research (interviewing industry experts)

Snowball and Quota Sampling

• Snowball sampling uses initial participants to recruit additional subjects
  • Particularly useful for hard-to-reach populations
  • Starts with a small group of known participants
  • Each participant refers other potential subjects from their network
  • Sample size grows like a rolling snowball, hence the name
  • Effective for studying hidden populations (undocumented immigrants)
  • Can introduce bias as participants tend to know similar others
• Quota sampling sets quotas for specific subgroups within the sample
  • Aims to create a sample that reflects known population characteristics
  • Researcher determines the proportion of each subgroup to include
  • Combines elements of stratified and purposive sampling
  • Often used in market research to ensure representation of key demographics
  • Does not involve random selection within quotas
  • Can lead to biased results if quotas are not properly set or filled