A brief introduction of methods
In term of designing of a research, authors must try to identify and formulate research questions based on several aspects. They could be obtained by solving the current trends or problems in certain topics, or by trying to understand the fundamental understandings that have not been answered yet. Although there are several perspectives and sources in formulating a research question, a researcher must be able to express three aspects of studies, i.e.,
- Research problems, including research questions and purposes
- Designing the research methods, including
- Sample collection methods
- Data collection methods
- Analysis methods
- Writing ability
These three, in my opinion, should be done in order. It means that without the understanding of research problems, a researcher would not be able to design the methods. However, in this article, the first step is considered to be done, and the focus is on sampling techniques.
Understanding of sampling techniques
One of the most crucial features in a research proposal is the suitability of the methods to the research problems. Problems occurred in this situation often happens in term of deciding the methods that are suitable. For instance, what kind of methods that are effective and efficient to understand the relationship between COVID-19 patients to their lifestyle activities in improving their autoimmune systems to prevent COVID-19 infections. Also, what is the public opinion about an increase fuel-price in certain area? Although quantitative studies appear to be arguably different to qualitative studies (sometimes incredibly different), one thing that is sure for both types of studies is the sample collection methods.
Considering the importance of samples selection, understanding the sampling techniques is a must. Thus, to create a good sample is firstly by identifying the population of samples to the study related, and this could also involve in many approaches, such as random and non-random sampling. Also, sometimes these two approaches are popular with the word probabilistic and non-probabilistic sampling, respectively.
What is a population?
Let us assume that a researcher is interested to identify diseases occurred in a small city in a country. This city has been censused, and the total of inhabitant is around 30,000 people distributed in 5 different areas. In this example, the people are divided into different types of clans, heights, weights, sexes, and any socio-geographical features. Different demographical features are well-defined collection of any individuals, shared in this city. In other words, a population in a study is individuals or objects that must have equal probability of being chosen.
As the purpose of our study is trying to identify various of illness, a study is conducted; however, taking 30,000 people would take a lot of time and energy. Therefore, the study must select the people to represent the purpose, and this is what we called as research sample. Remember, a research sample is selected to form an extrapolation that is used to gain generalization. The selection of this sample could be divided into five types:
- Random sampling
- Stratified sampling
- Systematic sampling
- Cluster sampling
- Convenient sampling
Please be noted, the order of these types showed the representational characteristics of samples. Thus, 1 is the highest representative characteristic, while the 5 is the least representative.
Random sampling
As it is mentioned previously, various demographical and socio-geographical characteristics might affect the purpose of our study. Taking certain samples based on these characteristics may be extrapolated in generalization. So, random sampling could be defined as taking samples from a population that has been identified based on the purpose of the study without specifying the criteria of the samples. For instance, a researcher wants to identify the illness in the study, so what are the ages, sexes, marital status, and any features, and this is defined as random sampling data. These data are generalized data, which demonstrate their generalizability. The question is, given that the data showed 100% in certain characteristic, such as 100% female, which doesn't show generalization; then, the samples could be bias due to the sampling error.
On the other hand, the continuation of this study appears to identify illnesses occurred, for instance, in elderly. Of course, to obtain generalization, representatives from each elderly communities (with the status of elder) must be collected based on the random sampling data. As a result, sexes and ages become the simple characteristics to collect the data; however, how many people does the researcher need? In this case, the use of random number generator software could be utilized in order to get the data, as we know surveying all of the people would spend a lot of budget and time. This technique is called as simple random sampling.
Stratified sampling
The word "stratify" is used to express an arrangement into strata. This means that the sampling takes place by stratifying the number of representative samples into a smaller number of samples. The reason behind this stratified approach is due to a high number of data that may affect our ability in collecting the sample. For instance, of total 30,000 population in the city that has 6 districts showed 3000 people (19% elders in district A), 7000 people (20% elders in district B), 4000 people (20% elders in district C), 6000 people (25% elders in district D), 5500 people (15% elders in district E), and 4500 people (1% elders in district F). Due to the very small number in a sub-district, it will not be represented in the sample (although it is impossible).
Therefore, stratifying the population before carrying the sampling technique is recommended. This means, we can stratify all the population, for instance, 10% from the total population, which is 3000 people. Then, identifying based on the percentages could be done, which are respectively selected in random stratified population into 570 elders (19%), 600 elders (20%), 600 elders (20%), 750 elders (25%), 450 elders (15%), and 30 elders (1%). In each district, random selection of the elders could be assigned as samples. Stratified sampling is considered probabilistic since it does include a random selection component due to the random chosen of members in each stratified district.
Systematic sampling
Systematic sampling is the most problematic in terms of generalizability. Using this technique means the sample group is identified by selecting random point within the population. Take a look example of the 30,000 participants of elders in the city, and the study selects the systematic sampling for 250 samples. To do this, the study must list every elder individual, and then pick a list-number randomly as starting list. Afterward, the study must choose the individuals started the list in every 250-turn. As a result, if the starting-list is in number 10, the sample is listed as 10, 260, 510, ..., until the samples account for 250 elders.
Cluster sampling
Given that the total population is 30,000 people which shows difficulties in surveying all people, a trued random sampling could be difficult to do. For instance, the study focuses investigating dietary characteristics (vegan and non-vegan diets) to the types of cardiovascular illness for elders in district C, in which elderly population showed 800 individuals, surveyed in 20 sub-districts (40 elders). These 800 elders become the study population, and the study is not allowed to randomly choose which individuals that can perform vegan and non-vegan; especially for those who needs protein. Other reason is due to the current conditions, such as physical or metabolic system of elders that cannot ethically be altered. Therefore, cluster sampling could overcome these two problems.
In order to perform cluster sampling, the study has decided a smaller sample size for 160 elders. Next, we could randomly select random sub-districts (among these 20 sub-districts), which accounts for 4 random sub-districts from the all-20 sub-districts. This cluster system could be considered as probabilistic data; however, limitation occurs, especially in the selected-random sub-districts may have had a lot of elders without any cardiac illness, which is not representative.
Convenience sampling
The convenience is often accidental, which recruit participants that just happen to be in a given place at a given time. This type of sampling is considered as non-random (non-probabilistic) sampling as the consequences of non-random sampling. The validity and generalizability of the results can be negatively affected. Aside the convenience sampling, there are examples of this technique, which are quota sampling, purposive and snowball sampling.
An illustration of this convenience sampling is a study that asks opinions of a doctor in a hospital about the importance of informed consent for performing all types of surgeries. The sample is convenient because these doctors work there, even though the generalization is limited in term of place, which is hospital. Subsequently, the importance of informed consent only originates from doctors who actually experience this situation. Although the argument could be made generalizable (for the doctors), opinion from another medical personnel, such as nurses, or any other staffs, as well as the patients is also important.
In conclusion, understanding of sampling techniques defines the results and discussion sections of the research proposal. Even more, the sampling techniques also relate to the methods of analyzing data, which latter answer the study hypotheses.
Tags
The Sciences