management assignment
1
Kong MGM 403
Sampling
2
Kong MGM 403
Population versus Sample
• Population (universe): complete group of entities that share some common set of characteristics – Depends on research
purposes
– People, stores, products
• Sample: subset of a populationPopulation Sample
3
Kong MGM 403
Population versus Sample • Study Wegmans’ shoppers behavior
– Population: people 18-65 yrs old, shop at any Wegmans store during 2017, spent more than $200 in 2017
– Sample: e.g., shop at Alberta Drive Wegmans store
• Study marketing research expenditure of Fortune 500 companies in 2016 – Sample: e.g., offices in NYC
• Study package information of premium bag coffee currently sold at Top 5 retail chains in U.S. – Sample: e.g., sold in NY state
• Study employee (staff) satisfaction of School of Management, UB
4
Kong MGM 403
What is Sampling
• Sampling: process of getting information from a sample – Census: getting information from all individual
elements of a population; usually very costly or impossible
• By studying a sample, we hope to make inference about characteristics of the population
5
Kong MGM 403
Learning Objectives
1. Learn how to develop sampling plan
2. Understand sampling error and determine sample size
3. Understand statistical property of sample characteristics
1. Define the Target Population
• Specify elements, extent, time
2. Choose the Data Collection Method
• Determine how you collect the sample - mail, Internet, telephone, mall intercept, etc. (discussed in survey design)
3. Select the Sampling Frame
• This is a concrete list for target population • You sample from this list
Developing a Sampling Plan
4. Select the Sampling Method
• Choose way to select elements (from the frame)
5. Determine the Sample Size
• Consider budget, accuracy needs
6. Execute the Sampling Plan
• Field workers must be trained to execute the sampling plan properly.
Developing a Sampling Plan
3. Select the Sampling Frame
• Sampling frame is a (concrete) list of members in target population
– E.g., list of Wegmans loyalty card members, telephone book, directory for firms in an industry, mailing list purchased from a commercial organization, Nielsen Consumer Panel
• There is seldom a perfect correspondence between the sampling frame and the population of interest
– Recognize and treat sampling frame error – Or redefine population in terms of sampling frame
3. Select the Sampling Frame
• Study consumer confidence of a city – Population: all households in the city – Sampling frame: the city’s telephone book
• Not include: households without telephone, and those with telephone but unlisted
• Unlisted households significantly different from listed
• Study Wegmans’ shoppers behavior
– Population: people 18-65 yrs old, shop at any Wegmans store in 2017, spent more than $200 in 2017
– Sampling frame: list of Wegmans loyalty card members
10
Kong MGM 403
Some Common Panel Vendors
• Some big names in online consumer panel providers:
• For a long list that can be geography-specific (and generally for marketing research/service firms):
– http://www.greenbook.org/market-research-firms.cfm/online-panels
• Costs per respondent varies by target population – For general population usually around $5-$15 but can be as
high as $75 or more for very targeted, hard to reach, or valuable individuals
• Error that occurs because the sample selected is not perfectly representative of the population.
• Decreases as sample size increases
• All error other than sampling error - also called “systematic error.”
• Result from nature of study design, correctness of execution, etc.
• Persists as sample size increases
Sampling Error
Nonsampling Error
Sampling and Nonsampling Error
4. Sampling Methods
13
Kong MGM 403
Sampling Methods
Probability Sampling
• Each element has a known, nonzero likelihood (probability) of being sampled
• Laws of probability hold and hence can calculate sampling error
Non-Probability Sampling
• Elements are selected based on some nonrandom manner (e.g., convenience)
• Cheaper than probability sampling
• Not feasible to infer sampling error
14
Kong MGM 403
Probability Sampling
• Simple random sample – All elements have the following chance of being
sampled
15
Kong MGM 403
Advanced Probability Sampling
• Cluster sampling (usually on geography)
– Randomly select a “cluster” then sample randomly within
– Can have multiple layers of clusters (e.g., metro areas- blocks-addresses)
– Purpose: reduce costs of collection
• Ideally each cluster should be a small-scale representation of the population
Geographic areas selected for national or regional surveys in progressively smaller population units, such as counties, then residential blocks, then homes.
16
Kong MGM 403
Advanced Probability Sampling
• Stratified sampling
– Divide original population into mutually exclusive and exhaustive subsets (strata)
• Choose strata (e.g., genders, income levels) that closely related to research interests. E.g., use income as strata in Wegmans shopper study
– Randomly sample within each strata
– Purpose: increase precision without increasing cost; include all important groups (especially minorities)
• Force the sample to be representative by making sure important dimensions of the population are represented in the sample
• Convenience sampling: A sample based on using people who are easily accessible – such as mall intercepts or other high traffic locations, or whatever is convenient
• Least expensive and least time consuming
• Judgmental sampling: The researcher selects who should be in the study based on personal judgement (because he or she believes that they are representative of the population or are otherwise appropriate).
• E.g., test markets selected to determine potential of a new product
Nonprobability Sampling
• Quota sampling: may be viewed as two-stage judgmental sampling. In first stage, researchers establish quotas based on demographic or classification factors. In second stage, sample elements are selected based on convenience or judgement
• Quotas ensure composition of sample is the same as composition of population regarding characteristics of interest
• Obtain representative samples at relatively low cost
• Snowball sampling: A sample in which additional respondents are selected based on referrals from initial respondents.
Nonprobability Sampling
20
Kong MGM 403
Quota Sampling: An Example
• A study about UB freshmen
• Suppose sample size is 200, the respondent quota would be
White Non-White
Male 40% 15%
Female 35% 10%
Categories Respondent quota
White male 80 (200*40%)
Non-White male 30
White female 70
Non-white female 20
Total 200
21
Kong MGM 403
Strengths and Weaknesses of Basic Sampling Techniques
Technique Strengths Weaknesses Nonprobability Sampling Convenience sampling
Least expensive, least time-consuming, most convenient
Selection bias, sample not representative, not recommended for descriptive or causal research
Judgmental sampling Low cost, convenient, not time-consuming
Does not allow generalization, subjective
Quota sampling Sample can be controlled for certain characteristics
Selection bias, no assurance of representativeness
Snowball sampling Can estimate rare characteristics
Time-consuming
Probability sampling Simple random sampling Easily understood,
results projectable
Difficult to construct sampling frame, expensive, lower precision, no assurance of representativeness Can decrease representativeness
Stratified sampling Include all important subpopulations, precision
Difficult to select relevant
stratification variables, not feasible to stratify on many variables, expensive
Cluster sampling Easy to implement, cost effective
Imprecise, difficult to compute and interpret results