Sample Design (Click icon for audio) Dr. Michael R. Hyman, NMSU
Photographic Example of How Sampling Works
Sampling Terminology Population or universe Population element Census Sample
Population/Universe Any complete group People Sales territories Stores Total group from which information is needed
Census Investigation of all individual elements that make up a population
Sample Subset of a larger population of interest
Stages in Selecting a Sample Define the target population Select a sampling frame Determine if probability or non-probability sampling method will be chosen Stages in Selecting a Sample Plan procedure for selecting sampling units Determine sample size Select actual sampling units Conduct fieldwork
Define Target Population Look at research objectives Relevant population Operationally define Consider alternatives and convenience
Select Sampling Frame List of elements from which sample may be drawn Mailing and commercial lists can be problematic (more on this later)
Sampling Units Group selected for the sample Can be persons, households, businesses, et cetera Primary sampling units Secondary sampling units
Choose Probability or Non-probability Sample Known, nonzero probability for every element Non-probability sample Probability of selecting any particular member is unknown
Conditions Favoring Non-probability vs. Probability Samples
Different Sampling Techniques
Non-probability Samples Convenience Judgment Quota Snowball
Convenience Sample Also called haphazard or accidental sampling Sampling procedure for obtaining people or units that are convenient to researchers
Discrepancy between Implied and Ideal Populations in Convenience Sampling
Judgment Sample Also called purposive sampling Experienced person selects sample based on his or her judgment about some appropriate characteristics required of sample members
Discrepancy between Implied and Ideal Populations in Judgment Sampling
Quota Sample Various population subgroups are represented on pertinent sample characteristics to the extent desired by researchers Do not confuse with stratified sampling (discussed later)
Representative Quota Sample Requirements
Snowball Sample Initial respondents selected by probability methods Additional respondents obtained from information provided by initial respondents
Probability Samples Simple random sample Systematic sample Stratified sample Cluster sample
Simple Random Sample Ensures each element in the population has an equal chance of selection
Systematic Sample A simple process Every nth name from list will be drawn
Stratified Sample Probability sample Sub-samples drawn within different strata Each stratum more or less equal on some characteristic Do not confuse with quota sample
Drawing a Stratified Sample: Example
Disproportionate Stratified Random Sampling Used by A.C. Nielsen
Cluster Sample Purpose: to sample economically while retaining characteristics of a probability sample Primary sampling unit is not individual element in population Instead, it is larger cluster of elements located in proximity to one another
Examples of Populations and Clusters
More Examples of Clusters
Strengths and Weakness of Sampling Techniques
Bases for Choosing a Sample Design Degree of accuracy Resources Time Advanced knowledge of population National versus local Need for statistical analysis
After Sample Design is Selected Determine sample size Select actual sample units Conduct fieldwork
Sampling Error
Types of Sampling Errors Sampling frame error Random sampling error Non-response error
Errors Associated with Sampling
Random Sampling Error Difference between sample results and result of a census conducted using identical procedures Statistical fluctuation due to chance variations
Key Aspects of Sample Frame Error
Systematic Errors Non-sampling errors Unrepresentative sample results caused by flawed study design or imperfections in execution rather than chance
Example Mailing Lists
More Mailing List Examples
Problems with Lists Representativeness Omissions and duplications Recency
Directories and Telephone Interviewing Directories not current Demographics and socioeconomics of voluntary non-list members differ from list members Solution Random digit dialing Add ‘1’ to listed number
Weighting Samples
Weighting a Sample
Internet Samples
Internet Sampling is Unique Internet surveys allow researchers to rapidly reach a large sample Survey should be kept open long enough so all sample units can participate
Advantages and Disadvantages Internet samples may be representative of target populations e.g., visitors to a Web site Hard to reach subjects may participate Major disadvantage Lack of PC ownership & Internet access among certain population segments
Web Site Visitors Unrestricted samples are clearly convenience samples Randomly selecting visitors Questionnaire request randomly "pops up" Over-representing more frequent visitors
Panel Samples
Panel Samples Typically yield high response rates Members may be compensated for time with sweepstake or small cash incentive Database on members Demographic and other information from previous questionnaires Select quota samples based on product ownership, demographics, lifestyle, or other characteristics
Recap Basic sampling terminology Stages in selecting a sample From target population definition to drawing the sample Non-probability vs. probability samples Types and appropriate usage Sampling error Internet and panel samples