startAbout UsFor ConsumersFor FoodserviceFor Educators


The source of one's data must be carefully selected.  If collecting data directly from people, this is called your sample.  A poor sample will result in poor quality data.  Sometimes one is limited in sample selection and so one must always describe how the sample was selected in the evaluation findings.


The population is a group of individuals, objects, or items having a common characteristic.  It is the total group from which a sample is selected.  Examples of a population might be all citizens in a county, all foodservice establishments in a county, or all children registered in an after-school food safety program.


A sample is a subset of the population.  A good sample is a smaller version of the population.  The best sample is representative of the population.  A sample is representative if important characteristics, such as age, educational level, ethnicity, income, are distributed similarly in both the sample and the population.  

For example, if one is collecting data about food safety practices in foodservice establishments in County A, the population is all foodservice establishments in County A.  Most of the time one cannot access the entire population.  In some counties, there are too many foodservice establishments, so it would not be feasible or even possible to study all.  It would cost too much money, time, and effort. Therefore, a sample must be selected.

Sampling is when one selects a smaller number (foodservice establishments) from the population (all foodservice establishments in the county) so an estimate can be made about the population (all foodservice establishments in the county).  A properly selected sample can result in data that is accurate and precise, whereas one that is not properly selected will yield poor results.

Selecting the Sample

There are two ways to select a sample –- random sampling and purposeful sampling.  Each is described below.

Random sampling. The sample is randomly selected from the population.  Every member of the target population can be included in the sample.  It eliminates subjectivity when choosing a sample.  Random sampling allows one to say that a sample is representative of the target population.  Following are frequently used random sampling methods.

  • Simple random sampling.  All individuals in the population have an equal chance of being selected.  One first needs a list of all members of the population.  This list is called the sampling frame.  Individuals on the list are selected one at a time.  Once an individual has been selected, there name is not returend to the selection pool. 
  • Systematic sampling. All members in the population are listed and every nth person is chosen after a random starting place is selected.  For example, one has a list of 450 foodservice establishments (the sampling frame) from their local health department. From this list a sample of 45 foodservice establishments must be selected for a study about food safety practices in foodservice establishments.  Divide 450 by 45. This yields 10. Select one out of every 10 foodservice establishments.  To systematically sample from the list, a random start is also needed.  Toss a die to get a number or pull a number from 1 to 10 out of a hat.  If the number 7 was selected, the 7th name on the list is selected first, then the 14th, 21st, 28th, until 45 names are selected.
  • Stratified sampling. To assure that certain subgroups are represented in proportion to their numbers in the population, subgroups are separately numbered and a random sample is selected. Examples of subgroups for a population of foodservice establishments might be schools, Fast Food restaurants, and family-style restaurants.  A clear rationale must exist when selecting any subgroup.  Stratified sampling is more complicated than simple random sampling.  Furthermore, using too many subgroups or "strata" can lead to a large and an expensive sample.
  • Cluster sampling. The unit of sampling is not the individual but rather a naturally occurring group of individuals, such as a school classroom, participants in several offerings of a training program, or the foodservice establishment.  Clusters are randomly selected, and all members of the selected cluster are included in the sample.  Cluster sampling is often used in large-scale evaluations involving surveys.
  • Matrix sampling. One sample receives one set of questions. Another sample receives a different set of questions.

Purposeful sampling. This type of sample is chosen based on the characteristics of the target population and the need of the evaluation.  Some members of the target population might have a greater chance of being chosen than do others.  A purposeful sample does not rely on random selection of units so findings are not viewed as representative.

One might have to choose a purposeful sample because an accurate list of the population might not be available, resources are limited, or obtaining cooperation from potential respondents is perceived to be difficult. A purposeful sample may be chosen to be sure to include a wide variety of people based on a number of critical characteristics.  Sometimes, individuals are specifically chosen to represent a specific characteristic.  More frequently, this type of sampling is used because it can be conveniently assembled.  Following are examples of purposeful samples:

  • Accidental sampling. This is the weakest type of sample, but is the easiest to get. "Man-in-the-street" interviews are typical of accidental samples.  One usually uses the first five or 10 people who happen along and are willing to talk.
  • Reputational sampling. This involves selecting specific people to respond to a survey or to be interviewed about an issue.  The choice of an individual depends on someone's judgment of who is and who is not a "typical" representative of the target population.
  • Convenience Sampling. A convenience sample is a group of individuals that is readily available for data collection.  For example, food safety training participants are chosen because of convenience.  This is probably the most common sampling method that is used in Extension.
  • Snowball Sampling. This type of sampling relies on previously identified members of a group identifying other members of the population.  As newly identified members name others, the sample snowballs.  This technique is useful when an accurate list of the population is not available.

Determining the Sample Size

Sampling error is large when the sample size is small.  Therefore, it is best to use as large a sample as possible.  The following table may be used to determine sample size base on a 5% error rate.


Population size Sample Size Population size Sample size
10 10 275 163
15 14 300 172
20 19 325 180
30 28 350 187
40 36 375 194
50 44 400 201
65 56 450 212
75 63 500 222
90 73 1,000 286
100 83 2,000 333
125 96 3,000 353
150 110 4,000 364
175 122 5,000 370
200 134 6,000 375
225 144 8,000 381
250 154 10,000 385
275 163 100,000 398

If the target population that is being evaluated has a total of 175 members, one needs to collect data from 122 members (175 is the population size and 122 is the sample size).  If the target population has 10,000 members, one needs to collect data from 385 members.  As the size of the population increases, it becomes easier to use a random sampling method because there are more individuals to choose from.

Test Your Knowledge

1.  What is the definition for population?

2.  What is a sample?  

3.  What are two ways that a sample can be selected?

4.  To assure that certain subgroups are represented in proportion to their numbers in the population what type of sampling method is used?

5.  If there are 325 restaurants in your county, what should the sample size be?