Primary Market Research

1
Primary Market Research

• Sampling

2
Module 4b Objectives

• Participants will
• define population, sample and sampling
• identify target population for PMR
• explain the rationale for choosing sampling
  method and size
• justify use of non-probabilistic sampling for
  PMR
• explain application of non-probabilistic sampling
  to PMR.

3
Sampling andPrimary Market Research

• Main input for product-planning Needs and
  expectations of Product customers
• Product customers end-users, clinicians,
  caregivers, and/or other stakeholders
• Primary market research captures relevant
  information from all of them

4
Target or Reference Population

• Group of people/objects that meet the criteria or
  have the characteristics relevant to research
  purpose
• For PMR All product customers that possess the
  information, knowledge and experience we are
  seeking plus other pre-defined criteria such as-
• low vision, wheeled chair user, left-handed,
  (human factors)
• working women, elderly over 65 living alone....
  ( demographics)

5
Population Types

• Homogeneous Population
• Group members with relevant characteristics are
  uniformly distributed throughout a Homogeneous
  population.
• Ex End-users of AugCom devices having similar
  experience and perspectives on use.

6
Population Types contd.

• Heterogeneous population
• Distinct sub-groups make up a Heterogeneous
  population relevant characteristics are
  uniformly distributed within sub-groups, not
  across sub-groups
• Ex Population comprised of users of AugCom
  devices, their caregivers and clinicians, all
  having related but different perspectives on
  device use

7
Heterogeneity vs. Homogeneity

• The information that people possess and which is
  sought by PMR Dependent or Outcome variable
• The variety in the information comes from
  differences between people on other
  characteristics age, experience. independent
  variables
• Ind. Variables gtdistinct subgroups gt
  Heterogeneity in the population

8
What is Sampling?

• Sample -part or subset of the whole group or
  target population that research is focusing on.
• Sampling - procedure by which to choose elements
  (people or objects) from the target population to
  make up a sample that has the same
  characteristics as the parent group.
• Purpose to describe or draw conclusions about
  the population through the sample, without having
  to study the entire group.

9
Sample Characteristics

• Sample data should let us confidently draw
  conclusions about population, so a sample should
  represent the target population characteristics.
  
• Representation is required in experiment-based
  research, because it allows accurate statistical
  generalizations about the population.
• In PMR, the reason for representation is to
  increase our credibility in using sample findings
  to describe the larger target population, rather
  than statistical generalizations.

10
Sample Characteristics

• In a sample representing a heterogeneous
  population subgroups will have the same relative
  frequency proportions as in the population with
  respect to relevant characteristics.
• Ex The number of wheelchair users, clinicians
  and caregivers in a mixed focus group sample
  should reflect their proportions relative to each
  other in the larger, mixed population

11
Sample Size is Important

• How many to include (sample size) is important
  for representing populations.
• Smaller the sample, more difficult to assure
  inclusion of members of the smaller sub-groups of
  population.
• Especially true of heterogeneous samples.
• Ex. Those who use specific features like switch
  scanning might not get into a small mix of AugCom
  users.

12
Restrictions on Sample Size

• Practical constraints reduce potential sample
  size
• Target population is reduced to accessible
  population, for not everyone is accessible for
  sampling. Ex. people in remote areas.
• Not everyone in the accessible population can or
  will participate.
• Not everyone selected will show up.

13
Sample size for PMR

• Experimental Research Statistical analysis
  methods define the minimum sample size required
  for accurate generalization.
• PMR Optimal sample size is guided by
• Information needs. Ex In designing this
  product, do we need broad coverage of all
  features or narrow, in-depth focus on specific
  features?
• Sample type. Heterogeneous samples needs to be
  bigger.
• Cost, logistics scheduling, availability

14
Sampling MethodsTwo Choices

• Random or probability sampling
• Non-probabilistic Sampling

15
Random orProbability Sampling

• The preferred way of Experiment-based research.
• Most reliable way (i.e., with least error) of
  generalizing from sample data
• No selector bias. Every person/object selected
  from the population has an equal chance of being
  included in the sample.
• Types Simple random, systematic, stratified,
  disproportional, cluster Portney Watkins,
  1993

16
Random Sampling Procedures

• Simple Random Start by choosing an element at
  random from the target population, continue to do
  so until the desired number of elements are
  selected for the sample. Use of a random number
  table is a good tool to draw elements
• Others also draw elements randomly from the
  target population, but a pre-defined condition
  modifies the drawing. See next slide.

17
Modified Forms of Random Sampling

• Systematic randomly draws every nth element
  from an organized target population. Ex. from a
  telephone directory, a dictionary of words
• Stratified randomly draws from sub-groups or
  strata. Ex. randomly choose 5 students from every
  classroom of a school
• Cluster/ multistage randomly draws pre-defined
  clusters of elements. Ex. Draw n schools
  from city schools, then m classes in each
  school. Others ..

18
Non-Probabilistic Sampling

• Population units have unknown probabilities of
  being included in the sample.
• Allows for selector bias
• Often a necessary alternative due to reality
  constraints cost, timeliness, sample size,
  access to target population,
• Types Convenience, Quota, Purposive, Snowball.

19
Purposive Sampling

• Researcher hand-picks people/objects
  purposefully allowing pre-defined
  characteristics/ criteria (Ex. special human
  factors) to be included in the sample.
• Its logic and power highly suit PMR research
  purpose more concerned with validly describing
  the sample and target population, than with
  statistical generalization.
• Often used successfully in qualitative
  evaluations.

20
Non-Probabilistic Sampling Other Forms

• Quota sampling pre-establishes inclusion of a
  certain quantity or quotaof elements in its
  sub-groups to represent the corresponding
  population subgroup characteristics
• Snowball or chain samples are built as the
  researcher carries out the selection process,
  getting referrals through sample members.
• Convenience Sampling includes elements based on
  availability Ex. every one that you can stop at
  a supermarket parking lot

21
Information Needs for PMR

• Context Product planning and development
• Sample data are used for -
• Formative Purpose - Data on needs and
  expectations guide designing decisions while
  product is still in development in the
  forming
• Summative Purpose - Data on product evaluation
  help end-of-the-development (disseminating/
  marketing) decisions.

22
Sampling Considerations for PMR

• PMR needs information both for Formative and
  Summative decisions
• PMR Samples should include
• -information-rich cases
• - preferably from every population sub- group.
  
• However..
• 3. Product customer universe is often
  heterogeneous with a considerable number of
  important subgroups.

23
Sampling Considerations for PMR

• In light of its information needs, using
  Probability Sampling for PMR might imply
• Either a small sample that excludes an important
  minority subgroup
• Or a sample of cost prohibitive magnitude that
  includes all important groups.
• Purposive sampling is a more useful alternative
  for constructing valid PMR samples of optimal
  size.

24
Sampling Considerations for PMR

• Useful alternatives
• Maximum variant sample mixed group with
  information-rich cases drawn from every subgroup
  of heterogeneous population.
• Ex Group of Hearing aid users, caregivers and
  clinicians
• b. Separate homogeneous samples of
  information-rich members for each subgroup
• Ex caregiver samples, user samples,
  manufacturer samples

25
Sampling Considerations for PMR

• c. Intensity samples include cases that
  intensely, but not extremely, manifest the
  information.Ex industry experts related to
  Wheeled mobility technology
• d. Random purposeful samples smaller random
  samples from a larger purposeful group.
  Increases credibility in generalizing not
  statistically to the target group
• e. Others Critical case, snowball Patton,
  1990

26
A Practical Sampling Alternative

• Combine purposive, quota and snowball sampling
  into your sampling rationale VIEW Example
• Before recruiting, prepare a Sampling frame or
  matrix to define how you will draw
  information-rich cases and distribute them in
  your sample.
• Define column and row headings by the different
  criteria (or characteristics) levels. Ex columns
  to represent physical ability levels (high and
  low) to operate an AAC device, rows for
  environmentaldemands (high and low) on device use

27
A Practical Sampling Alternative contd

• 3. Define quotas or optimal numbers of people
  to fill the cells with, after weighing the
  corresponding proportions known or estimated of
  target population subgroups against reality
  (time, cost and logistical) constraints
• 4. Fill each cell purposively with the desired
  numbers by recruiting people that meet criteria
  as defined. Use Snowball strategy for
  recruitment,if necessary.

28
Where Do You Use Samples in PMR?

• PMR collects information through
• -focus groups interviews
• -surveys
• -one-on-one or telephone interviews

29
Recruitment

• Sampling frame defines what and how many specific
  types of people you want to include
• Recruitment implements the selection
  plan. -contact individuals
• -get commitment
• -schedule and logistics

30
Recruitment Challenges

• Quite often, not everyone approached by recruiter
  meets the criteria, and not everyone that meets
  the criteria is readily identifiable or
  accessible.
• Use the Snowball approach. Get people through a
  chain referral process to fill in the pre-set
  sampling frame.
• This adds the snowball rationale to the
  purposive-quota rationale begun at the sample
  planning stage.

31
Recruitment Guidelines

• Define sampling matrix first and then select
  people by recruiting. Plans for criteria,
  population characteristics, number, etc. should
  precede recruitment, so rational adjustments can
  be made when the plan cannot be fully achieved.
• Over-sample- allow for bigger proportion of
  underrepresented segments
• Over-recruit - counteract sample attrition
  anticipate logistical, scheduling problems.
• Recruitment takes time - start early

32
Sampling for PMRAn Example

• The attached example of sampling protocols for
  the caller-connect device illustrates the
  foregoing rationale for focus group interviews

33
Sampling protocols The Case of the Caller
-connect Device

• Purpose of the Focus Group interviews To obtain
  information useful for Concept Refinement
• features/characteristics of a device that meets
  the need of people that leave telephone off the
  hook for various reasons stress, functional
  limitations, cognitive impairment, forgetfulness
  by older and child family members

34
Sampling protocols The case of the Caller
-connect Device

• Step one define target population
• driving question is What features should make
  up this "off-the-hook" device?
• seeks input for a "universal design"
• universe to include expertise from specific
  "groups" e.g. families with children/elderly
  leaving phone off the hook with various
  functional needs and with relevant demographics.
  
• Basically, purposive sampling makes sense.

35
Sampling protocols The case of the Caller
-connect Device

• Step two make a sampling plan or chart and
  define what proportions to include
• hearing "all" subgroups of interest impractical
  Alternatively, define several independent subsets
  of universe and then draw a sampling chart for
  each subset
• 3 groups defined -- persons with disabilities,
  elderly, younger adults with children

36
(No Transcript)
37
(No Transcript)
38
Related Issues

• Questions, Comments, Suggestions?