|
Marketing Research &Information:
Lecture 7 Sampling
Aaker, Kumar, Day (2007) Ch 14
Bradley (2007) Ch 5
Saunders, Lewis & Thornhill (2007)
Research Methods for Business Students
The Research Process
Identify problem/info need
Research objectives
Decide on your approach
(methodology), and formulate a plan
Research Method &
Data Collection Tool
Sample Method
Collect information
Analyse the data
Present the findingsSampling
n Choosing the type of people to survey
n Locating them
n Deciding how to select the individuals
involved
n Deciding how many to survey
n Identifying how representative the
information collected isStages of sampling
1. Examine the objective of the study (PURPOSE)
2. Define the people of interest (POPULATION)
3. Find suitable source for the population members
4. Decide on the sampling type and approach
(PROCEDURE)
5. Decide on the sample size
6. Proceed with the fieldwork
7. Correct sampling errors ready for reporting
(PUBLICATION)What is sampling?
n The process of obtaining information from a
subset of a larger group
n Population
n The entire group of people (universe) about whom
info is needed
n Census
n A collection of data obtained from every member
of the population
n Sample
n A subset / segment of all the members of a
population of interestSampling
n For what reasons might it be necessary
to select a sample rather than collect
data from the population?Sample Design
Probability
sample
n A member has a
positive, calculable
probability of being
chosen
n Response rate
important
n Is expensive
n Objective
Non-probability
sample
n Uses human judgment
n Subject to errors that
cannot be measured
n Usually cheaper than
random methods
n SubjectiveSampling Methods
n Probability
n Simple random sampling
n Systematic sampling
n Stratified random sampling
n Cluster sampling
n Non-probability
n Convenience sampling
n Judgement sampling
n Quota sampling
n Snowball samplingSampling Process
1. Define the population of interest
2. Determine whether to sample or
census
3. Select the sampling frame
(probability only)
4. Determine sample size
5. Choose a sample method
6. Implement1. Define the population of
interest
n Specify the characteristics of the individuals
from whom or about whom information is
needed to meet research objectives
n Bases for defining the Population of Interest
n Geographic area
n Demographics
n Usage
n Awareness2. Determine whether to sample
or census
n Difference between a sample and a
population?
n Choice depends on size of population,3. Select the sampling frame
n Sampling frame is a list of population
items/elements/ units. The sample is
selected from this list.
n OR the frame is a specific procedure for
generating such a list
n The population must be defined – e.g.
production employees; car owners
n NB take care when using phone
directory as a listGood frames are
n up to date
n complete
n affordable
Poor frames are
n old
n incomplete
n inappropriate
Sampling FramesCommon Sources
Common frames include
n the Electoral Register
n the Postcode Address File (PAF)
n Telephone directories
n Subscriber / membership records
n Customer recordsIdentify a suitable sampling
frame for each of the following:
n A study to identify how company directors of
manufacturing firms where there are >500
employees think a specified piece of
legislation will affect their companies?
n A study to identify how employees at
Cheltenham Gardens Ltd think the proposed
introduction of compulsory Saturday working
will affect their working lives?How sample size is
determined
n Sample size is generally agreed before
fieldwork, but in some projects it may be
determined during fieldwork
n The sample size for any study depends on
n the purpose, in terms of required precision
n the size and nature of the population under study.
n Procedural aspects such as time, budget &
resources available
n Publishing aspects, in terms of importance of
resultsHomogeneous?
The important thing is
how homogeneous
that population is
rather than its full
size
For 1 million people
who are identical we
only need a sample of
one person. If there are
200 different people we
need to interview all
200.Sample size decisions
Main ways to decide the sample size:
n by calculation
n by using "accepted" industry
standards
n by budget (time or money available)
n by "building" analysis cells5. Determine sample size
n Factors in selecting sample size:
n Variability in the population
n Required level of confidence
n Required limits of accuracy
n Likely level of non-response
n Requirements for analysis of subgroups
n Practical factors – e.g. budgetsDetermine sample size
n Governed by:
n Confidence levels
n Desired margin of error
n Types of analyses planned
n Size of population (see end of notes for table
of sample sizes)
n Expected response ratesExample calculation of
response rate
n Estimate the sample size required, e.g.
210
n Out of the questionnaires returned (e.g.
129, deduct the voids, e.g. 8).
Responses total 121
n 121 as a percentage of 210 = response
rate, i.e. 57.6%Example calculation of sample
size when response rate is known
n If you estimate that you will get 25% of
your questionnaires returned, and you
know that the desired sample size for
the population is 360, the actual sample
you should contact is:
360 x 100 = 1440
25Probability sampling
n Probability sampling- every member of the
population has a known likelihood of selection
n Therefore can only be used when some sort
of a list identifying members of the
population is available, e.g. staff list;
customer list; accounts (as in customer
accounts) database
n AKA representative samplingProbability sampling methods
n Simple random sampling
n free from bias
n uses a sampling frame, sample generated by
generating list of random numbers
n Systematic sampling
n Choose every nth item after a random start
n Advantages
n Easy to use and cost effective
n Disadvantages
n Could be biased if there is a regular pattern to the
population which coincides with the sampling method
n Not completely randomExample calculation of a
systematic sample
n There are 750 people employed by an
organisation. With a 5% confidence level, 254
employees should comprise the sample. What
is the sample fraction (i.e. how many names
should be skipped each time the sample is
being selected from the list of employees)?
750/254 = every third name should be
selected (approx)
n see Saunders, M. Lewis, P. & Thornhill, A.
(2007:212)Probability sampling methods
n Stratified random sampling
n Variation on random sampling. The
population is divided into strata /
categories. Random samples taken from
each strata.Probability sampling methods
n Cluster sampling: non-random sampling
method that involves selecting one definable
subsection taken to be representative of the
population in question
n Advantages
n Good alternative to stratified sampling if
satisfactory sample frame not in existence
n Inexpensive to operate
n Disadvantages
n Potential for considerable biasProbability sampling
n Advantages
n Sampling error can be calculated
n Results can be projected to the whole
population
n Researcher can be sure of obtaining results
from a representative cross section of the
whole population
n Disadvantages
n Expense, time consumingNon-Probability sampling
n Non-probability sampling – samples
where specific elements of the
population have been selected in a nonrandom
manner
n Used where there is lack of an available
frame to select the sample from, e.g.
coffee drinkers.
n The process is otherwise similarNon-Probability sampling
methods
n Convenience sampling – where samples are
drawn at the convenience of the interviewer
n Advantages
n Able to gather substantial amounts of data from
high traffic locations
n Quick & easy to implement
n Disadvantages
n Potential for considerable bias
n Misrepresentation of populationNon-Probability sampling
methods
n Judgement (purposive) sampling – where the
interviewer uses an educated guess as to
who should represent the population
n Advantages
n Viable for focus groups
n Useful if researcher has knowledge about the
population
n Disadvantages
n Subjective, some members of the population will
have smaller chance of selectionNon-Probability sampling
methods
n Quota sampling – interviewers are told
to interview all the people they meet up#p#分页标题#e#
to a certain quota
n Advantages
n Cheap and easy, no sampling frame necessary
n Disadvantages
n Bias – cannot make a valid estimate of
sampling errorQuota Sampling
Most common
method,
typically uses
n Age
n Sex
n Social GradeTypes of Quota
Interlocking
n (interrelated)
n Strata are linked
n Greater control
over interviewer
n Difficult task
n Representative
Non-interlocking
n (independent)
n Strata unconnected
n Simple task
n Likely to skew to
co-operative people
n Less usedNon-Probability sampling
methods
n Snowball sampling – new respondents
selected based on referrals from initial
respondents
n Advantages
n Good for low incidence populations,
reduces research costs
n Disadvantages
n Bias – respondents may be reluctant to
give referralsNon-probability sampling
n Advantages
n Cheaper & quicker
n Acceptable where accuracy is not critical –
i.e. exploratory research
n Disadvantages
n Unaware as to what extent the sample
represents the population
n Sampling error cannot be calculated, so
results cannot be accurately projected onto
the whole populationn Qualitative research
always uses nonprobability
methods of
sampling
n So subjectivity and
human judgement is
always involved
Qual & quant
sampling differencesRepresentative?
n Qualitative samples will NOT be
representative of the population
n Many quantitative samples ARE
representative of the population
n But some are drawn to examine extremes,
non-users, or users of above-average
quantities. Samples may be selected to
understand minorities.
n The end-sample may be representative of
these groups, but not the overall populationSample sizes
n In Qualitative research sample size
determination is more subjective
n In theory sample sizes should not be fixed
firmly at the start of the project. The
overriding idea is that new cases should be
selected until the data brings nothing new.
n In practice a methodological compromise is
made and most proposals set a certain
number of groups and depths. This then
allows budgets to be controlled and
timetables to be developedSources
n Aaker, D. Kumar, V. & Day G. (2007)
Marketing Research.9th Edn. New
Jersey: Wiley
n Bradley, N. (2007) Marketing Research
Oxford:OUP
n Saunders, M. Lewis, P. & Thornhill, A.
(2007) Research Methods for Business
Students, Harlow: Pearson Education
|
