Sampling Design Process PDF

# SAMPLING DESIGN ## Sampling Design Process (Schindler, 2022) The sampling design is a subprocess of research design. It involves the following steps: 1. Define the target population and a case (entities [collectively and individually] possess the desired information about the specific variables...

# SAMPLING DESIGN ## Sampling Design Process (Schindler, 2022) The sampling design is a subprocess of research design. It involves the following steps: 1. Define the target population and a case (entities [collectively and individually] possess the desired information about the specific variables and parameters). 2. Define the population parameters of interest (summary descriptors [proportion, mean, variance] of study variables) in the population. 3. Identify and evaluate the sample frame (list of cases within the target population). 4. Define the cases needed (choose between a census or sample; select any sample size). 5. Define the appropriate sampling method (the type of sample to be used). 6. Define the sampling selection and recruitment protocols (choose standardized or custom-design procedures). ## Target Population and Case (Schindler, 2022) In business, a target population can be any of the following, with a case being a single element drawn from that target population: - *People.* These are individuals or groups (i.e., employees, customers, suppliers). - *Organizations or institutions.* These are companies, trade associations, professional online communities, unions, etc. - *Events and happenings.* These refer to trade association meetings, presentations to financial analysts, industry conventions, employee picnics, etc. - *Objects or artifacts.* These pertain to products, machines, production waste or byproducts, tools, maps, process models, ads, etc. - *Settings and environments.* These refer to warehouses, stores, factories, distribution facilities, etc. - *Texts.* These pertain to annual reports, productivity records, social media posts, emails, memos, reports, etc. **Example:** The researchers at Metro University (Metro U) are exploring the feasibility of creating a dining club whose facilities would be available on a membership basis. They will need to make a substantial investment to launch this venture, but the research will allow them to reduce many risks. **Research Question:** Would a membership dining club be a viable enterprise? **Investigative Questions:** - Who would patronize the club, and on what basis? - How many would join the club under various membership and fee arrangements? - How much would the average member spend per month? - What days would be most popular? - What menu and service formats would be most desirable? - Given the proposed price levels, how often would each member have lunch or dinner per month? Should the target population be defined as "full-time day students on the main campus of Metro U?" or should it include "all persons employed at Metro U" Or should townspeople who live in the neighborhood be included? Without knowing the likely patron for the new venture, it is not apparent which of these is the appropriate target population. Assume the Metro U Dining Club is solely for the students and employees on the main campus. The researchers might define the population as "currently enrolled students and employees (full- and part-time) of Metro U, main campus, and their families." In the context of feasibility studies, the target population represents the target market (potential customers who will purchase the proposed product or service). ## Population Parameters (Schindler, 2022) Population parameters are summary descriptors (e.g., incidence proportion, mean, variance, etc.) of variables of interest in the population. Sample statistics are descriptors of those same relevant variables computed from sample data. Sample statistics are used as estimators of population parameters and are the basis of inferences about the population. Table 1 indicates population parameters for the Metro U dining study. | Population Parameter of Interest | Data Level and Measurement Scale | |---|---| | Frequency of eating on or near campus at a restaurant within the last 30 days | - Ratio data (actual number of eating experiences). - Ordinal data (less than five [5] times per month, more significant than five [5] but fewer than 10 times per month, greater than 10 times per month). | | The proportion of students/employees expressing interest in the dining club | - Nominal data (interested, not interested). | | The proportion of students/employees spending money per person per visit | - Interval data (P50 to P99, P100 to P149, P150 to P199, P200 to 249). | *Table 1. Example Population Parameters in the Metro U Dining Study. Retrieved from Business Research Methods (14th ed.), 2022, p. 93* Depending on how measurement scales are set, each collects a different data level, generating different sample statistics. Thus, choosing the parameters of interest will dictate the sample type and size. Data have different properties depending on how they were collected. Table 2 reviews the data types and these properties (the entire discussion was covered in Week 1). | Data Type | Characteristics | Exam ple | |---|---|---| | Nominal | Classification | - Respondent type (faculty, staff, student). | | Ordinal | Classification and order | - Preferred doneness of steak (well done, medium well, medium rare, rare). | | Interval | Classification, order, and distance | - The rating of the last restaurant experience (scale of 1-10; 1 = inferior, 10 = exceptional). | | Ratio | Classification, order, distance, and natural origin | - The average amount spent per person for the last dinner in a restaurant. | *Table 2. Review of Data Types and Characteristics. Retrieved from Business Research Methods (14th ed.), 2022, p. 94* When the variables of interest in the study are measured on interval or ratio scales, the sample mean is used to estimate the population mean, and the sample standard deviation is used to calculate the population standard deviation. When the variables of interest are measured on nominal or ordinal scales, the sample proportion of incidence (p) is used to estimate the population proportion and the pq to estimate the population variance where q = (1 − p). The population proportion of incidence is "equal to the number of cases in the population belonging to the category of interest, divided by the total number of cases in the population." Proportion measures are necessary for nominal data and are widely used for other measures. The most frequent proportion measure is the percentage. ## Sample Frame (Schindler, 2022) The sample frame is the list of cases in the target population from which the sample is drawn. Ideally, it is a complete and correct list of population members only. However, as a practical matter, the sample frame often differs from the desired population. The Metro U directory would be the logical first choice for the dining club study as a sample frame. Another way to make the sample frame for the Metro U study would be to secure a supplemental list of the new students and employees and a list of the withdrawals and terminations from Metro U's registrar's office and human resources database. Sample frames can also be drawn from a community. Drawing participants for research from a pre-recruited virtual community is a common practice as it has been shown to increase the response rate, increase completion rate, speed data collection, lower the cost of research, and gain deeper insights with its ability to combine qualitative and quantitative methodologies. Using a community as a sample frame increases as more business managers conduct agile or iterative research to test ideas. For feasibility studies, a researcher may use census data from the local municipality where the business will be established as a sample frame. From the census data, the percentage of the population not included as the target market shall be eliminated using the market demand analysis framework. It deals with the demand conditions of the proposed business venture. To determine the net potential market, the researcher must determine the total population in the target business location. The population not qualified as consumers must be eliminated from the total sample frame based on predetermined characteristics (i.e., age, gender, race, etc.). **See sample computation below:** Household population in the City: 110,721. Population who are not qualified as consumers: 170/1. ## Number of Cases (Schindler, 2022) The ultimate test of a sampling design is how well any measured cases represent the target population's characteristics. ## Sample Versus Census Most people intuitively understand why drawing a sample works. One taste from a drink reveals whether it is sweet or sour; it is unnecessary to drink the entire glass. Selecting a few ads from a magazine is assumed to reflect the characteristics of the complete set. If some staff members favor a particular software training strategy, it is inferred that others will also. The basic concept of sampling is that conclusions about the entire target population can be drawn by selecting some cases from a population. There are several compelling reasons for using a sample (a subset of the target population) rather than a census (all cases within a population), as follows: 1. **Greater Speed of Data Collection.** Due to the smaller number of cases in a sample, using a sample drawn from a target population will always take less time than conducting a census. 2. **Availability of Population Cases.** Some situations require using a sample. Safety is a compelling appeal for most vehicles, yet evidence is needed to make safety claims. Therefore, cars are crash-tested to evaluate bumper strength or the efficiency of airbags in preventing injury. In testing for such evidence, the vehicles tested are destroyed. A census would require the destruction of all cars manufactured. 3. **Better Quality Results.** W. Edwards Deming suggests samples possess "the possibility of better interviewing (testing), more thorough investigation of missing, wrong, or suspicious information, better supervision, and better processing than is possible with complete coverage." Error related to research comes from two (2) sources: errors related to the sample itself (sampling error or estimates of a variable drawn from a sample differ from the actual value of a population parameter) and errors not related to the sample but to all other decisions made in the research design (non-sampling error). ## Sample Size Assuming that a sample was chosen over the census, how many cases should comprise the sample? The most pervasive myths are (1) a sample must be significant, or it is not representative, and (2) a sample should bear some proportional relationship to the size of the population from which it is drawn. With non-probability samples, researchers use subgroups, rules of thumb, and budget considerations to settle on a sample size. In probability sampling, how large a sample should be is a function of the variation in the population parameters under study and the estimating precision needed by the researcher. Some principles that influence sample size include: - The greater the dispersion or variance within the population, the larger the sample must be to provide estimation precision. - The greater the desired precision of the estimate, the larger the sample must be. - The narrower or smaller the error range, the larger the sample must be. - The higher the desired confidence level, the larger the sample must be. - The greater the number of subgroups of interest, the greater the sample size since each subgroup must meet minimum sample size requirements. Cost considerations influence decisions about the size and type of sample and the data collection methods. Probability sample surveys incur list costs for sample frames, callback costs, and various unnecessary costs when nonprobability samples are used. Research has budgetary constraints, which may encourage a researcher to use a nonprobability sample. When the data collection method is changed, the amount and type of data that can be obtained also change. ## Sampling Method Probability and non-probability sampling are the two (2) main types of sampling methods. In probability sampling, respondents are randomly selected to participate in a survey or other research method. For a probability sample, each person in a population must have an equal chance of being selected, and the researcher must know the probability that an individual will be chosen. Probability sampling is the most common form for public opinion studies, election polling, etc. (Elliot, 2020). Non-probability sampling is when a sample is created through a non-random process. It requires a researcher to send a survey link to their friends or passersby on the street. Non-probability samples are often used during the exploratory stage of a research project and in qualitative research, which is more subjective than quantitative research but are also used for research with specific target populations (Elliot, 2020). The four (4) primary methods of probability sampling are as follows (McCombes, 2023): | Method | Description | Example | |---|---|---| | Simple Random Sampling | In simple random sampling, every member of the population has an equal chance of being selected. The sampling frame should encompass the entire population. | To select a simple random sample of 1,000 employees from a social media marketing company, assign a number from 1 to 1,000 to each employee in the company database. Then, use a random number generator to select 100 numbers. | | Systematic Sampling | Each member of the population is listed with a number, and instead of randomly generating numbers, individuals are chosen at regular intervals. | List all employees alphabetically. Randomly select a starting point from the first 10 numbers (i.e., number 6). Select every 10th person from that point (i.e., 6, 16, 26, etc.) until a sample of 100 people is formed. Ensuring that no hidden pattern in the list (e.g., grouping by team and seniority) might skew the sample is essential. | | Stratified Sampling | Stratified sampling involves dividing the population into subgroups (strata) based on relevant characteristics to ensure proper representation and draw precise conclusions. - Divide the population into subgroups (i.e., gender, age, income, job role). - Calculate the number of people to be sampled from each subgroup based on their proportions in the population. - Random or systematic sampling will be used to select individuals from each subgroup. | If a company has 800 female and 200 male employees, and the sample needs to reflect the company's gender balance, sort the population into two (2) strata based on gender. Random sampling was used on each group to select 80 women and 20 men, resulting in a representative sample of 100 people. | | Cluster Sampling | Cluster sampling divides the population into subgroups with characteristics similar to the whole sample. Instead of individuals, entire subgroups are randomly selected. - Divide the population into subgroups (clusters). - Randomly select entire clusters. - If clusters are large, sample individuals within each cluster (multistage sampling). This method is helpful for large, dispersed populations but has a higher risk of error due to potential differences between clusters, making it challenging to ensure representativeness. | A company with offices in 10 cities, each with a similar number of employees in similar roles, wants to collect data without traveling to every office. By randomly selecting three (3) offices, these become the clusters from which data is collected. | The following are non-probability sampling techniques (McCombes, 2023): | Method | Description | Example | |---|---|---| | Convenience Sampling | A convenience sample includes the individuals who are most accessible to the researcher. | A researcher is gathering opinions about student support services at their university. After each class, they ask fellow students to complete a survey. Although this method is convenient for data collection, it results in a sample that is not representative of all students at the university, as it includes only those taking the same classes at the same level. | | Voluntary Response Sampling | Like a convenience sample, a voluntary response sample is mainly based on ease of access. Instead of the researcher choosing participants and directly contacting them, people volunteer themselves (i.e., by responding to a public online survey). Voluntary response samples are always at least somewhat biased, as some people will inherently be more likely to volunteer than others, leading to self-selection bias. | A researcher sends a survey to all students at their university, and many students decide to complete it. While this can provide some insight into the topic, the respondents likely strongly oppose student support services. Consequently, their opinions may not be representative of the entire student body. | | Purposive Sampling | This type of sampling, also known as judgment sampling, involves the researcher using their expertise to select the most helpful sample for research purposes. It is often used in qualitative research, where the researcher wants to gain detailed knowledge about a specific phenomenon rather than make statistical inferences or where the population is tiny and specific. An adequate purposive sample must have clear criteria and rationale for inclusion. Always describe your inclusion and exclusion criteria, and beware of observer bias affecting your arguments. | A researcher wanted to understand more about the opinions and experiences of disabled students at a university. Purposefully, he selected several students with different support needs to gather varied data on their experiences with student services. | | Snowball Sampling | If the population is hard to access, snowball sampling can be used to recruit participants via other participants. The number of people you have access to "snowballs" as you contact more people. The downside here is representativeness, as you cannot know how representative your sample is due to the reliance on participants recruiting others. It can lead to sampling bias. | The researcher investigated the experiences of homeless people in their city. Probability sampling wasn't possible because there was no comprehensive list of all homeless people in the town. They met one person who agreed to participate in the research, and she put them in contact with other homeless people she knew in the area. | | Quota Sampling | Quota sampling relies on the non-random selection of a predetermined number or proportion of units, known as a quota. The researcher divides the population into mutually exclusive subgroups (called strata) and recruits sample units until the quota is reached. These units share specific characteristics determined by the researcher. | The researcher aims to gauge consumer interest in Boston's new produce delivery service, focusing on dietary preferences. They divide the population into meat eaters, vegetarians, and vegans, drawing a sample of 1000 people. To ensure the company can cater to all consumers, they set a quota of 200 people for each dietary group. This approach ensures that all nutritional preferences are equally represented in the research, allowing for easy comparison between the groups. The researcher continued recruiting until they reached the quota of 200 participants for each subgroup. | Feasibility studies commonly use a simple random sampling technique to determine the target respondents since it is the most efficient way to draw a sample population. The minimum number of respondents must be at least 30; an acceptable maximum number is 10% of the net potential market. - Household population in the City: 110,721 - Population who are not qualified as consumers (7%): 7,750 - Net Potential Market: 102,971 - Target no. of respondents: 10% = 10,297.10 The researcher can trim the target respondents based on the minimum acceptable number. Using the simple random sampling method, the researcher may select respondents that represent the entire scope of the business venture (i.e., if the city has 10 barangays, ... ### References: - Elliot, R. (2020). Probability and Non-Probability Samples. https://www.geopoll.com/blog/probability-and-non-probability-samples/ - McCombes, S. (2023). Sampling Methods. https://www.scribbr.com/methodology/sampling-methods/ - Schindler, P. (2022). Business Research Methods (14th ed.). McGraw Hill.