SAMPLING PROCEDURES

1
SAMPLING PROCEDURES

• MKTG 3342
• Fall 2008
• Professor Edward Fox

2
SAMPLING PROCEDURES

• Outline
• 1. INTRODUCTION Sampling vs. Census
• 2. PROCEDURE FOR DRAWING SAMPLE
• 3. TYPES OF SAMPLING PLANS
• 4. NONPROBABILITY SAMPLES
• 5. PROBABILITY SAMPLES

3
1. SAMPLING VS. CENSUS

• ADVANTAGES OF SAMPLING
• Time
• Cost
• DISADVANTAGES
• Because only a sample has been drawn, there is
  associated uncertainty (error)

4
1. SAMPLING VS. CENSUS
Sample
5
2. PROCEDURE FOR DRAWING SAMPLE (slightly
different from procedure in your book)

• A. DEFINE THE TARGET POPULATION
• the specification of people or cases on whom the
  research is to be conducted.
• B. IDENTIFY THE SAMPLING FRAME
• Listing of population elements from which sample
  is drawn.
• C. SELECT THE SAMPLING PLAN
• D. DETERMINE SAMPLE SIZE
• E. SELECT THE SAMPLING UNITS

6
3. TYPES OF SAMPLING PROCEDURES

• SAMPLE DESIGN
• NONPROBABILITY SAMPLES PROBABILITY SAMPLES
• - CONVENIENCE - SIMPLE RANDOM
• - JUDGMENTAL - STRATIFIED
• - QUOTA PROPORTIONATE
• - SNOWBALL DISPROPORTIONATE
• - CLUSTER
• - SYSTEMATIC
• 
  

7
PROBABILITY VS. NONPROBABILITY

• PROBABILITY SAMPLING
• Every member of the population has a known,
  non-zero probability of being selected
• NON-PROBABILITY SAMPLING
• The probability of any particular member being
  chosen for the sample is unknown

8
NONPROBABILITY SAMPLING METHODS

• CONVENIENCE SAMPLES
• Nonprobability samples used primarily because
  they are easy to collect
• JUDGMENT SAMPLES
• Nonprobability samples in which the selection
  criteria are based on personal judgment that the
  element is representative of the population under
  study

9
NONPROBABILITY SAMPLING METHODS

• QUOTA SAMPLES
• Nonprobability samples in which population
  subgroups are classified on the basis of
  researcher judgment
• SNOWBALL SAMPLES
• Nonprobability samples in which selection of
  additional respondents is based on referrals from
  the initial respondents

10
PROBABILITY SAMPLING METHODS

• SIMPLE RANDOM SAMPLING
• A probability sample in which every element of
  the population has a known and equal probability
  of being selected into the sample

Sample Size
Probability of Selection
Population Size
11
PROBABILITY SAMPLING METHODS

• STRATIFIED RANDOM SAMPLING INVOLVES THE FOLLOWING
  TWO-STEP PROCEDURE
• I. The parent population is divided into
  mutually exclusive and collectively exhaustive
  subsets (strata)
• II. A simple random sample is chosen from each
  subset

12

• REASONS FOR STRATIFIED SAMPLING
• -- Investigate characteristics of interest by
• subgroup stratification allows for
• adequate representation of different
  subgroups
• -- Increase precision (reduce sampling error)
• EXAMPLE
• Suppose I want to investigate if low-income users
  default more on credit card than high-income
  users. I want to ensure adequate representation
  of people with both high and low incomes, so I
  divide the population on the basis of income and
  take a random sample from the high-income group
  and the low-income group.

13

• PROPORTIONATE VS. DISPROPORTIONATE STRATIFIED
  SAMPLING
• PROPORTIONATE STRATIFIED SAMPLING Take sample
  size in (same) proportion to size of the
  population in each subgroup or stratum e.g.,
  suppose there are 3,000 high-income users and
  10,000 low-income users then take maybe 30 (1)
  high-income and l00 (1) low-income users
• DISPROPORTIONATE STRATIFIED SAMPLING Sample
  size not necessarily in proportion to population
  subgroup size e.g. take 60 (2) high-income
  consumers and 100 (1) low-income users because I
  think there is substantial variation among
  high-income consumers

14

• CLUSTER SAMPLING
• TWO-STEP PROCEDURE
• -- Population is divided into mutually
• exclusive and collectively exhaustive
  subsets
• -- A random sample of the subsets is selected
• -- In one-stage cluster sampling, all elements
  in
• the randomly selected subsets are included
• -- In two-stage cluster sampling, a sample is
• selected probabilistically from each randomly
• selected subset

15

• MOTIVATION FOR CLUSTER SAMPLING
• GENERALLY LOWER COST (but less accurate)
• For example,
• In the income / credit default case, suppose you
  divide people based on where they live (say, by
  zip code), then randomly select zip codes (say
  75248 and 75212) and investigate either everyone
  in both zip codes or a random sample of people
  from both zip codes

16

• DIFFERENCE BETWEEN STRATIFICATION AND
  CLUSTERING
• The variable used for stratification must be
  related to research focus (income, in our
  example)
• The variable used for clustering must not be
  related to research focus (zip code, in our
  example)

17
PROBABILITY SAMPLING METHODS

• SYSTEMATIC SAMPLING
• Probability sampling in which the entire
  population is numbered. The first number is drawn
  randomly. Subsequent elements are drawn using a
  skip interval.

Population Size
Skip Interval
Sample Size
18
PROBABILITY SAMPLING METHODS

• Example of systematic sampling
• Suppose I want to pick 100 phone numbers to call
  from a telephone directory with 1000 pages. Use

Population Size (1000)
10
Skip Interval
Sample Size (100)

• First, draw a random number between 1 and 10 (say
  you get 7) then pick pages 7, 17, 27, 997
• From each page you can pick a phone number (say
  on top right corner)

19
SUMMARY OF KEY POINTS(1 of 2)

• The population is the total group of people in
  whose opinions one is interested
• A census involves collecting desired information
  from all the members of the population
• A sample is simply a subset of a population

20
SUMMARY OF KEY POINTS(2 of 2)

• Probability sampling methods are selected in such
  a way that every element of the population has a
  known, nonzero probability of selection
• Nonprobability sampling methods include all
  methods that select specific elements from the
  population in a nonrandom manner
• Stratified probability sampling is generally the
  best method for selecting a sample, if time and
  budget permit