Sampling

1
Sampling
2
Sampling

• To do most research, one must have people to
  study.
• Sampling refers to selecting cases, or plain and
  simple, getting a group of people (or other
  elements) out of the population to study.
• Whenever we attempt to make statements about a
  set of people in general using a smaller group of
  peoplegeneralizingthe data we use is from a
  sample.
• Sample vs. Census
• Census A complete count of an entire
  population
• So why dont we always do a census?

3
Sample vs. Population
Sample
Population
4
Sampling

• Types of Samples (You can sample almost
  anything)
• Case Studies Persons in Field Studies Contexts
  Observed
• Archival Data Experiment Participants
• Persons answering a Survey
• Depending on how the sample was generated, there
  are limits on how much we may generalize.
• Given limits on generalizability, the purpose of
  your research will help determine the type of
  sampling you do.

5
Sampling

• Sampling Techniques
• Nonprobability Sampling methods that do not let
  us know in advance the likelihood of selecting
  for the sample each element or case from a
  population
• vs.
• Probability Sampling methods that allow us to
  know in advance how likely it is that any element
  of a population will be selected for the sample
• Knowing the chance of selection allows one to
  control sampling bias (under or
  overrepresentation of a population characteristic
  in a sample)

6
Sampling

• Sampling Techniques
• Nonprobability
• (Very common in psychology, medicine, sociology)
• Availability Sampling, convenience
  samplingselection of cases based on what is
  easiest to get
• Experiments
• Exploratory and Qualitative research
• Avoid this if you can
• Quota SamplingKnowing something about your
  target population, you select your availability
  sample to ensure that it looks similar to your
  population

7
Sampling

• Sampling Techniques
• Nonprobability
• Snowball SamplingRespondent-driven sampling
  where initial respondents refer others to the
  researcher
• Usually used with hard-to-discover populations
• Bias introduced by structured nature of
  affiliation
• Can be improved with incentives to subjects to
  recruit a certain number of new respondents
• Purposive Samplingtargeting select people for a
  sample because of their unique position
• Helps get understanding of systems or processes
  or information on a target population
• Not representative of population in general

8
Sampling

• Sampling Techniques
• Nonprobability
• Nonprobability samples have limited
  generalizabilityyou can never be sure the sample
  represents the population
• But, researcher can work to establish what the
  sample represents
• Why use nonprobability samples?
• Well-suited for exploratory and evaluation
  research
• Nonprobability does not mean intentional attempt
  to make sample nonrepresentative
• We cannot all be identified by sampling frames,
  sometimes making nonprobability sampling more
  accurate
• More Efficiency
• Social and social psychological processes can
  be effectively studied and described
• No project is ever enough anyway, community of
  scholars can add information through other
  researchcollections of projects can create a
  complete picture

9
Sampling

• Sampling Techniques
• Probability Sampling Sampling methods that
  allow us to know in advance how likely it is that
  any element of a population will be selected for
  the sample
• Goal A representative sample of a target
  population
• Probability sampling begins with a sampling
  frame, or a list of all elements or other units
  containing the elements in a population.
• E.g., Phone book, All Universities, Known
  Addresses, Subscribers to a magazine.
• If a sampling frame is incomplete (which they
  usually are) then the accuracy of the sample is
  compromised. The researcher has the burden of
  assessing the sampling error or bias.

10
Sampling

• Sampling Techniques
• Probability
• Simple Random Samplingcases are identified
  strictly on the basis of chance.
• Random number table to select from sampling frame
• Random digit dialing
• Equal probability of selection
• Systematic Random Samplingusing a list, the
  first case is selected randomly, then subsequent
  cases are selected at equal intervals.
• Typically the same as Simple Random Sampling
• Be aware of periodicity

11
Sampling

• Sampling Techniques
• Probability
• Cluster Samplingused when sampling frames of
  individuals are difficult to obtain, but clusters
  are identifiable. Randomly select clusters, then
  use the clusters sampling frames to select
  cases.
• E.g., There is no national list of independent
  Baptists, but almost all independent Baptist
  churches can be identified.
• Select down to smaller number of clusters, then
  do the difficult work of identifying elements
  (persons to participate)
• Generally better to maximize the number of
  clusters and minimize number of cases from each
  cluster because clusters tend to be homogeneous
• Often called multistage sampling. When one uses
  two or more successive sampling steps one is
  doing multistage sampling.
• Each stage produces sampling error more stages,
  more error

12
Sampling

• Sampling Techniques
• Probability
• Stratified Random Samplingsampling frame is
  divided into strata of interest, cases are drawn
  from each stratum on the basis of chance.
• Small subpopulations of interest may yield too
  few cases in simple random sampling. To
  compensate, the researcher draws samples from
  each subpopulation independently.
• E.G., Latino population of Santa Clara County is
  around 25. A random sample of 100 would produce
  20 30 Latinostoo few to generalize to Santa
  Clara County Latinos.
• Do independent sampling from each stratum.

13
Sampling

• Sampling Techniques
• Probability
• Stratified Random Sampling
• Proportionate Stratified Samplingselect cases in
  a way that ensures the same proportion from each
  stratum in the sample as exists in the
  population.
• Population 4 black, 25 Latino, 27 Asian, 44
  white
• Sample of 1,000 40 black, 250 Latino, 270
  Asian, 440 white
• Disproportionate Stratified SamplingProportion
  selected from each stratum is not the same as in
  the population.
• Population 4 black, 25 Latino, 27 Asian, 44
  white
• Sample of 1,000 250 black, 250 Latino, 250
  Asian, 250 white
• Idea is to get a lot of cases in each stratum
• When combining all cases into one sample, use
  weighted averages

14
Sampling

• Sampling Techniques
• Probability
• Just because a sample is random, that does not
  mean that it is representative or that the
  research is good.
• Limited Sampling Frame
• Think of presidential phone polls
• Who is at home? Type of person, day of polling,
  etc.
• Who has a land line?
• Problems of non-responserandom non-response
  okay, but systematic non-response is biasing
• Phone surveys typically do not report response
  rate. They are often below 30
• How were questions worded Measurement error
• Problems of misspecified models Leads to not
  asking the right questions

15
Sampling

• Sampling Techniques
• Probability
• Is the Sample large enough?
• Larger samples produce less sampling error
• Too large is a waste of money
• Big is good, but accurate and appropriate are
  better
• Fraction of population sampled does not increase
  accuracy unless fraction is very large
• The more heterogeneous the population, the larger
  the sample needed.
• The more variables of interest, the larger the
  sample needed.
• The weaker the effects, or the smaller the
  differences between groups, the larger the sample
  needed to see effects or differences between
  groups.
• TO SUM MORE COMPLEXITY REQUIRES LARGER SAMPLES