214-216

Systematic Sampling

  • Systematic sampling improves convenience over simple random sampling.

  • Potential issues with the order of elements in the sampling frame can often be resolved easily.

Stratified Sampling

Definition and Purpose

  • Stratified sampling organizes the population into homogeneous subsets to ensure representation.

  • It addresses the necessity of having proper numbers of each subset present in the sample.

Implementation

  • Possible stratification variables include class, gender, and GPA.

  • Example: Ensuring the sample includes correct numbers of freshmen, juniors, etc.

Benefits

  • Allows for a greater degree of representativeness and decreases sampling error.

  • Stratification improves the representativeness of the sample regarding age and other related variables.

Factors Reducing Sampling Error

  • Larger samples reduce sampling error compared to smaller samples.

  • Homogeneous populations produce samples with smaller sampling errors.

  • Example: 99% agreement on a statement yields low sampling error, 50-50 split increases sampling error significantly.

Choosing Stratification Variables

  • Variables should be relevant to the research and related to what is being represented accurately.

  • Common variables: age, race, and gender often correlate with crime-related factors.

  • Geographic area can enhance representativeness.

Representation in Stratified Samples

  • Stratified sampling ensures that different groups are adequately represented, enhancing the reliability of the results.

Disproportionate Stratified Sampling

Definition

  • The technique enables the collection of non-representative samples to focus on rare attributes of the population.

Purpose

  • Important for sufficiently representing rare cases by oversampling them.

  • Example: National crime surveys may oversample urban areas due to higher crime rates.

Examples

  • British Crime Survey oversampled areas served by smaller police forces to accurately represent rural regions.

Multistage Cluster Sampling

Definition

  • Used for populations without exhaustive lists for sampling, involving the selection of clusters of elements and then sampling within those clusters.

Application

  • Useful in large populations (e.g., all U.S. police officers) to create practical sampling units.

Steps

  1. Select primary sampling units (e.g., cities, counties).

  2. Obtain lists of elements within those units (e.g., law enforcement officers).

  3. Sample those lists for participant selection.

Steps in Multistage Sampling

Listing and Sampling

  • Involves iteratively listing and sampling units at different stages.

  • A sample of city blocks leads to listing residents, resulting in a sampled population.

Guidelines for Multistage Cluster Sampling

  • Maximize clusters while minimizing elements within each cluster to balance efficiency and accuracy.

  • Maintain a goal of selecting five households per census block as a general practice in population research.

Loss of Accuracy in Multistage Sampling

  • Sampling error can accumulate at each stage of the process, reducing overall sample accuracy.

  • Each stage contributes to increased potential sampling error compared to a singular, random sample.