Statistical Sampling

Statistical Sampling

• Convenience sampling
• Simple random sampling
• Systematic sampling
• Stratified sampling
• Cluster sampling
• Multi-stage cluster sampling

Convenience Sampling

• Easy, fast, inexpensive, and members are readily available

• Convenience samples are basically useless in statistics as they may not be representative of the population that it was drawn from--there is bias.

Simple Random Sampling

• Less complicated than other good sampling methods.
• In most cases, sampling bias is avoided if sample is drawn properly.

• The sample may not be good if you don't have full access to a large population to draw from (i.e. limited sampling frame)

Systematic Sampling

1. Start at a random starting point
2. Draw members at regular intervals through the population

• Quick and easy

• If individuals are sorted in any way (e.g. alphabetical) and only the first few individuals are selected, then this systematic sample will not be representative of the population.

Stratified Sampling vs. Cluster Sampling

1. Divide the population into homogenous, non-overlapping subgroups (i.e. strata) based on their characteristics
2. Randomly select members from each stratum.

• This reduces sampling variability and provides a better picture of the population.
• We have control of ratios of strata. This is called proportional allocation.

• Could be expensive and time-consuming

1. Divide the population into clusters of evident subgroups (e.g. geographical locations) that the population contains
2. Randomly select a few clusters.

• Usually cheap, quick, and easy.
• You can also get a larger sample size than if you do a SRS.

• Not very representative of the population since you are drawing from clusters where members in them have similar characteristics.
• Also, there can be a high sampling error, especially when some clusters are not accessible, leaving a large proportion of the population unrepresented.
• The sample will only be unbiased if each cluster is rather representative.

Two-Stage Cluster Sampling

Steps for two-stage cluster

1. Divide the population into clusters of evident groups that the population contains
2. Randomly select a few of those clusters.
3. Instead of including all members from each cluster in the sample, you perform SRS (or Systemic Sampling) on each of the selected clusters to draw members, and only those members get surveyed.

• More accurate and more random than cluster sampling with the same sample size.

• Difficult or tedious to do.