Sampling Methods in Research

Sampling Methods

• Sampling is the process of selecting observations.
• It's often not possible or necessary to collect information from everyone.
• Researchers select a smaller subset of observations and generalize the findings to the entire population.

Learning Objectives

• Understanding probability sampling allows representation of large populations with smaller subsets.
• The quality of a sample is determined by how well it represents the population.
• Probability sampling ensures every member of the population has a known, non-zero chance of selection.
• Probability sampling methods lead to representative samples.
• Sample statistics, grounded in sampling distributions and probability theory, allow estimating population parameters.
• Simple random sampling is the foundational probability sampling technique.
• Different probability sampling designs (systematic, stratified, multistage cluster) can be used.
• The National Crime Victimization Survey (NCVS) and the British Crime Survey employ multistage cluster sampling.
• Probability sampling is statistically more representative than non-probability sampling.

Introduction

• Sampling involves selecting a smaller group from a larger population.
• Collecting data from everyone isn't always possible or practical.
• Sampling helps researchers generalize findings from a subset to the wider population.

The Logic of Probability Sampling

• Probability sampling enables researchers to generalize findings about a select group to a larger unobserved population.
• Each member in the population has a known and equal chance of inclusion in the sample.
• Samples must reflect the variations within the population to be representative.

Conscious and Unconscious Sampling Bias

• Researchers should be aware of potential bias.
• A biased sample does not fully represent the larger population.
• Equal Probability of Selection Method (EPSEM) ensures representative samples.
• Samples with characteristics similar to the population are representative.

Sampling Distribution

• Sample element: individual or unit of study.

• Population: entire group of interest.

• Population parameter: value of a variable for the population.

• Sample statistic: summary description of a variable in the sample.

• Sample statistics are used to estimate population parameters.

• Purpose of sampling: selecting elements in a way to accurately portray population parameters.

• Random selection is crucial.

• Sampling distribution: range of sample statistics obtained from many samples.

• Sampling frame: list of all elements in the population used for random selection.

• Increasing the number of samples and interviewees increases the range of possible estimates from sampling.

Estimating Sampling Error

• Random samples from populations generate sample statistics dispersed around the population parameter.

• Probability theory provides formulas for estimating the closeness of sample specifics to a population parameter.

• Standard error measures sampling error.

• It indicates how the sample statistics distribute around the population parameter.

Confidence Levels and Intervals

• Sampling error accuracy is expressed using confidence intervals.
• Confidence intervals indicate a specified interval where population parameters likely lie.
• Confidence levels help determine the appropriate sample size for a study.

Discussion Questions

• What is a 100% confidence level response for survey results?
• What is the decision if choosing between NCVS or the British Crime Survey?
• What are the strengths and weaknesses of snowball sampling for research?

Nonprobability Sampling

• Nonprobability sampling methods might be used when impossible to collect a probability sample.
• Probability that an element is included in the sample is not known.
• Cannot generalize to larger populations (e.g, purposive, quota, snowball).

Nonprobability Sampling, cont.

• Purposive sampling: selecting samples based on the researcher's judgment and study purpose.
• Quota sampling: creating a sample reflecting population characteristics.
• Snowball sampling: identifying participants through referrals from initial participants.

Multistage Cluster Sampling

• Multistage cluster sampling involves selecting from stratified clusters.
• May be used when creating an exhaustive list of the population is impossible.

National Crime Victimization Survey

• This survey aims to represent the U.S. population aged 12 and older living in households.
• Primary sampling units (PSU) are automatically included, based on size and other characteristics.
• Census enumeration districts (CED) are selected, and clusters of four housing units are selected.

British Crime Survey

• The survey uses 289 parliamentary constituencies.
• The selection is stratified, considering geographic area and population density.
• Two sample points are selected and further divided into four segments for different addresses.
• One segment is randomly selected, and disproportionate sampling is used for inner-city residents.
• Eligible individuals are identified and selected by interviewers.

Populations & Sampling Frames

• Various probability sampling designs can be used in different research purposes.
• Key aspect: relationship between population and sampling frame.
• Sampling frame: quasi-list of elements from which a probability sample is drawn.

Simple Random Sampling

• Sampling frame elements are numbered, and random numbers determine sample inclusion.
• Forms the basis for probability theory and estimating population parameters.

Systematic Sampling

• Elements in a list are systematically selected for inclusion in the sample.
• A random start is crucial to avoid bias.
• Periodicity in the list can create bias.

Stratified Sampling

• Stratified sampling ensures appropriate numbers from homogeneous subsets within a population.
• Method creates more representative samples.
• Disproportionate stratified sampling allows capturing rare cases more effectively.