PROBABILITY AND NON-PROBABILITY SAMPLING TYPES (EXAMPLES) - CULTURE VOCABULARY - 2025

The types of sampling are the various ways of extracting data from a part of the total, a powerful statistical tool whose function is to determine what part of the population or the universe is necessary to examine, to make inferences and obtain information about it.

Sampling is very important when you cannot or do not want to analyze the entire population. Note that the term "population" does not refer only to a large group of people or living beings, but in general to the total of elements that are going to be studied in a given problem.

Figure 1. Sampling is important to select a representative sample from a universe. Source: Pixabay.

According to the type of sampling chosen, the part of the population that is considered the most representative is selected, always in accordance with the objectives.

Of course, when only part of the universe of data is taken, it is possible to miss some details and omit information, which is why the results will not be as accurate as they should be. This is known as sampling error.

The idea is to simplify the universe of data as much as possible, choosing the most representative sample that is capable of providing the maximum information, to ensure the validity of the results.

Types of probability or random sampling

A probability sampling is based on the probability that the subjects of the sample have to be selected. In this way, each element of the population is given a known chance of being chosen, which of course must be greater than 0.

This is extremely important, because it may happen that from a universe of data, a sample has been selected that is not sufficiently representative of the whole.

If so, the results will be biased, as some parts of the population will be more favored over others. To avoid bias, of which there are several categories, one option is to let chance select the sample and thus give each element a non-zero probability of being selected.

Simple random sampling

This is a simple way to ensure that chance does its job. For example, if it is a question of selecting some children in a class to participate in a school art event, all the children's names are placed on identical folded ballots, mixed in a hat, and a handful drawn at random.

All the children in the class make up the population, and the handful of ballots that were pulled out of the hat is the sample.

The success of the procedure lies in making a complete list of all the children, so that no one is left out. In a small course this is not a problem; But when you want to select a sample from a larger population, you have to refine the method.

Simple random sampling can be carried out with replacement or replacement. For example, if we extract some element from the population and return it after selecting and examining it, the universe of our elements always remains the same throughout the study.

If, on the contrary, the chosen element is studied, more is not returned, it is sampling without replacement. This must be taken into account when calculating the probability of an element of being selected.

Systematic random sampling

To carry out this sampling, it is also necessary to list N elements and also determine the size of the sample, which we will call n. The list is called a sampling frame.

Now the jump interval is defined, which is denoted by the letter k and is calculated like this:

A random number is chosen - at random - between 1 and k, called ro random start. This is the first individual on the list to be selected and from there the following elements on the list are chosen.

An example: suppose you have a list of 2000 students from a university and you want to obtain a sample of 100 students to participate in a congress.

The first thing to do is find the value of k:

Once we have divided the total number of students into 100 fragments of 20 students, one of the fragments is taken and a random number is chosen between 1 and 20, for example 12. Therefore the twelfth student on our list is the random boot.

The next student to be selected must be 12 + 20 = 22, then 42, then 62 and so on, until all 100 are completed.

As you can see, it is a quick method to apply and that usually gives very good results, without having to put the 2000 names in a hat and take 100 of them, as long as there are no periodicities in the population, which give rise to biases..

Stratified random sampling

Figure 2. In stratified random sampling, the population is divided into segments called strata. Source: Pixabay.

In simple random sampling, each item in the population has the same probability of being selected. But this may not always be true, especially when there are more complexities to consider.

To carry out a stratified random sampling scheme, the population must be divided into groups with similar characteristics. These are the strata. The strata are then taken and simple random samples are chosen from each, which are then combined to form the final sample.

The strata are determined before sampling, studying the characteristics of the data universe.

These characteristics can be marital status, age, place where you live, for example urban, suburban and rural population, profession, level of education, gender and many more.

In any case, it is expected that the characteristics of each stratum will be very distinctive, that is, that each stratum is homogeneous.

Within the stratified sampling we distinguish two categories, according to whether the sample size of each stratum is or is not proportional to its size.

Random cluster sampling

The methods described above select the elements of the sample directly, but in cluster sampling, a group of elements is chosen from the population and these will be the sampling unit, which is called a cluster.

Examples of clusters are the departments of a university, geographic entities such as provinces, cities, counties or municipalities, all of which have the same probability of being selected. In case of choosing a geographic entity, we speak of sampling by areas.

Once the clusters have been chosen, the elements to be analyzed are chosen from there. Therefore, the procedure can have several stages.

This method has some similarities with the stratified random method, except that here some clusters are selected from the total, while in the previous method all strata of the population were studied.

Non-probability sampling types

Probability sampling may be very expensive in some situations, as time and resources must be invested to find samples that are truly representative.

It also often happens that there is not a complete sampling frame –the list-, therefore it is not possible to determine the probability of selecting an element.

For these cases, non-probability sampling types are used, with which information is also obtained, although there is no guarantee of precision in the results.

When this type of sampling is applied, some criteria must still be followed at the time of selection, seeking that the sample is as adequate as possible.

Convenience sampling

It is a fairly elementary type of sampling, in which the elements of the sample are chosen according to their availability, that is, selecting the individuals who are most at hand. It has the advantage of being a very low cost method, due to its speed and convenience.

But as said, there is no certainty of getting reliable information on your results. It is sometimes used to make quick, short polls before an election, or to inquire about customer preferences for certain products.

For example, a pollster can go to the exit of three of the shopping centers that are closest to his house and ask those who leave for which candidate they would vote. Or a teacher can survey their own students, because they have immediate access to them.

Although it looks as if the results of such a procedure are worthless, it happens that they could be a good reflection of the population, as long as there are good reasons to assume that the bias is not very large.

However, it is not so simple, because the students of a certain teacher may not constitute a representative sample of the rest of the student body. And most of the time, pollsters in shopping malls tend to interview the most attractive-looking people.

Quota sampling

To sample by quotas, a good prior knowledge of the population strata must be had, in order to have an idea of ​​which are the most representative elements. But it is not governed by the randomness criterion of stratified sampling.

In this type of sampling it is necessary to set "quotas", hence the name of the method. These quotas consist of gathering a number of elements with certain conditions, for example 15 women whose age is between 25 to 50 years, who do not smoke and also own a car.

Once the quota is determined, the first people who meet the established conditions are chosen. The criteria for this last step may be at the investigator's convenience. Here you can see the difference with the stratified sampling method, which is random.

However, it is a low-cost method that is advantageous if, as we said, the population under study is well known.

Snowball sampling

The procedure to follow in this style of sampling is to select a few people who lead others, and these in turn to others, until the sample is the size that the researcher needs.

This is a procedure that can be useful to characterize some populations with quite specific traits. Examples: inmates in a prison or people with certain diseases.

Discretionary sampling

Finally here it is the researcher who decides the criteria to be used to choose his sample, according to his knowledge. It can be useful when it is necessary to add certain individuals to the study, who, using a random method, could not participate.

References

1. Berenson, M. 1985. Statistics for Management and Economics, Concepts and Applications. Editorial Interamericana.
2. Statistics. Sampling. Recovered from: encyclopediaeconomica.com.
3. Statistics. Sampling. Recovered from: Estadistica.mat.uson.mx.
4. Explorable. Cluster sampling. Recovered from: explorable.com.
5. Moore, D. 2005. Applied Basic Statistics. 2nd. Edition.
6. Netquest. Probability sampling: stratified sampling. Recovered from: netquest.com.
7. Wikipedia. Sampling. Recovered from: es.wikipedia.org