Sampling

1
Sampling

• BUAD 259
• Business Research Methods
• Penwell

2
Sampling Definitions

• Population element -- A single example drawn from
  the population, e.g. you
• Census -- An investigation or study involving all
  elements within a population, e.g. survey all
  college students around the entire world.
• Sampling -- The process of using a small subset
  of a population to draw conclusions about the
  population as a whole.
• Population -- The complete set of some target of
  investigation, e.g. all college students
• Sample -- A subset of a larger population, e.g. a
  random selection of Mary Washington College
  Students.

3
Why Sample?

• Makes research plausible
• Efficiency
• Costs
• Time
• Reduce non-sampling error
• Accuracy
• Only option when destructive testing used

4
Sampling Frames

• Target Population -- The entire population that
  is focus of your investigation (College Students)
• Sampling Frame (aka - Working Population) -- List
  of elements from which to draw your sample (MWC
  Students)
• Sampling frame error - over or under
  representation of some set of elements of the
  population (e.g. excluding BLS JMC)
• Sampling Services -- providers of mail lists,
  phone lists, reverse directories, etc.

5
Sampling Frames Continued

• Sampling Units - single element within a sample
• Primary Sampling Unit (PSU) - selected in the
  first round
• Secondary or Tertiary Sampling Unit - selected in
  the second or third round

6
Error

• Random Sampling Error -- difference between the
  sample results and census results if identical
  procedures are used. Caused by chance variation.
• Systematic (non-sampling) Error -- due to design
  or execution, e.g. sample bias
• Non-Response Error -- Statistical differences
  between those who responded and those who did not
  respond

7
Non-Probability Sampling

• Non-probability sampling -- selected on basis of
  personal judgment or convenience, e.g. select
  your friends.
• Convenience (aka, haphazard, accidental,
  captive)
• Judgment (aka purposive, bellwether)
• Quota Sample -- Assure representation of all
  subgroups
• Snowball -- Built off an initial probability
  sample

8
Probability Sampling

• Probability Sampling -- Everyone in the
  population has some known probability of being in
  the selected sample, e.g. 1 in 20.
• Simple Random -- picked from a hat, or random
  number table
• Stratified -- Random selection within set strata
• Proportional or disproportional (to population)
• Systematic -- Select every 200th
• Periodicity - Bias due to nature of the list or
  interval
• Cluster -- Sampling unit is larger than the
  individual element of interest.
• Area Sampling -- randomly select representative
  cluster or randomly select elements within each
  cluster.
• Ideally, as heterogeneous as the population.
• Multistage -- Progressive application of
  probability sampling to find smaller subsets
  (National surveys)

9
Sampling Design Criteria

• For what purpose?
• Nature of the population
• Accuracy -- needs may vary
• Confidence intervals
• Generalizability -- needs vary by the nature of
  the study
• Costs and Time frame -- expenses and deadlines
• Tolerance for biases and error
• Statistical Analyses?
• See Exhibit 16.10