Strengths and weaknesses in sampling
Firstly, it is essential to understand a sample, and its purpose. A sample can be defined as a section of a population who are selected to be participants in a study. The specific selection of participants is chosen to give an overall representation of the whole population. Due to a variety of factors, particularly money and time constraints, it is not always possible to study the entire population, with the sample taking a considerable amount time, that when the sample if complete, the data acquired may no longer be representative of the population. As a result a sample is conducted, considered to be part of the population which is observed. (Cochran, 1977)
Without sufficient forms of sampling, generalising with a respectable degree of accuracy is unachievable. There are 2 types of sampling, non-random and random, and this sampling is reliant upon the notion of unsystematic selection. The basic principles denote that in order to provide generalisations related to unique population, the characteristics of the sample must reflect the characteristics of the unique population that has participated in the sample. Thus, a sample can be understood as a miniature population. The only accurate alternative would be to select the entire population to take part in the sample. Whilst there are scenarios where this could be achieved, it is highly unlikely that this could be achieved, but for the population being small enough.
When conducting a sample, many see a direct relationship between the overall accuracy of the sample and the population used to create the sample. Consequently, several people will consider a sample to be more accurate when the population used within the sample is greater. This is not necessarily true, as a sample of 100,000 people will not be 10 times as accurate as a sample with a population of 10,000. When conducting a sample, it is essential to consider a variety of factors when calculating the size of the sample that will be used. For example, the cost of the sample, the time duration of the sample, and the size of the population that will be used in order to obtain relevant information and the level of sampling error that will occur once the results of the sample are complete.
However, using a larger population when conducting a sample leads to less sampling error, also known as standard error. Therefore, this simply insinuates that the larger the sample, the smaller the error. As a result, those chosen for national samples and national surveys are assiduously selected, resulting in specific samples of only 2,000-3,000. With participants for samples and surveys being specifically selected, the level of sampling error diminishes considerably; nevertheless it is imperative to remember that sampling error can never be eliminated, irrespective of population size. (Barnett, 1991)
Random sampling, also referred to as probability sampling, involves a type of random selection which is responsible for choosing the element of the sample. Considerably more confidence can be found in random sampling compared to non-random sampling. The main methods when conducting random sampling include cluster, simple random, stratified random and systematic. The selection procedure ensures each element within the population has an equal, as well an independent chance of being selected to take part in the sample.
The elements within a non-random sample are selected through non-random method. This has a detrimental effect on producing representative samples compared to random sampling. However, many researchers still choose to use non-random samples when conducting their research. Their non-random samples are determined by the 3 main methods used within the sample; convenience, purposive and quota.
Random samplingensures that each and every member within the population has an equal and identical chance of being included within a sample. Thus, many believe random sampling to be the easiest, fastest and simplest method in order to draw a sample from a population. When choosing random sampling to conduct research, it is essential to have a complete and full list of the population in order to select a completely random sample. (Jessen, 1978) However, this can be seen as very difficult to accomplish. Developing a thorough population list is considerably simpler when using a distinct and smaller population. Several researchers and those involved in conducting samples consider random sampling to be most beneficial, believing random sampling should be used as often as possible. This is very much due to random sampling highlighting authentic, realistic and reliable generalisations. For example, researchers would prefer to conduct a random sample of 100 people, rather than a non-random a sample of 1,000 people, therefore highlighting random sampling as a general preference amongst those who are responsible for conducting samples. Accordingly the advantages of using random sampling are that the population sample is only influenced by chance, ensuring the sample is fair, non-biased and non-discriminative. However, obtaining a list of the entire population is complicated, and as a result this can prevent entirely random sampling. (Wiley, 1992)
The method of systematic sampling consists of two factors that will determine involvement in the sample, and they are chance and ‘the system’. This ‘system’ can be described as the process of facilitating random selection within systematic sampling. For instance, when selecting a sample of 50 names from the population of 100, instead of random selection taking place within the population to determine who takes part in the sample, a researcher may select every second name from within the population to take part in the sample. Systematic sampling can as similar to random sampling, unless a systematic bias is evident through the presentation of names on the population list. However, it is very simple to avoid a systematic bias from occurring, through examining the list prior to conducting the sample, as well as communicating with those in charge of putting together the list, asking how the list was assembled. (Foreman, 1991)
Quota sampling is a type of sampling that is frequently used in market research and in opinion polls. Those conducting the sample are given a quota of specified subjects to recruit. For example, when conducting a sample concerning favourite music, the interviewer might be asked to find and select 10 adult females, 10 adult males, 10 teenage boys and 10 teenage girls. (Wiley, 1992) However, many imperfections exist when conducting a quota sample, with the simplest fault being that the sample is not random; consequently this means that the sampling distributions of all and any statistics are unknown. The non-random sample is selected in a precise and specific manner in order to ensure that the known characteristics correspond with the overall population sample. When conducting a quota sample, it may be advantageous to set the quotas before the sample is selected; in spite of this it is also possible to use quota sampling strategies spontaneously. Some researchers may feel that it is not beneficial to carry out research prior to conducting the sample as they are working with an intact group. As a result, the researcher may include questions concerning the characteristics of his respondents; this is in addition to questions related to the outcome variables. The additional questions to those within the questionnaire should concentrate on the topics that are most expected to introduce biases. Once the data has been has been analysed, in order to validate that there are no obvious biases, the researcher could compare and contrast the characteristics between the population and the sample. For example, a small association with a minor budget may want to conduct a private investigation to find out the attitudes, opinion and viewpoints of British University students concerning alcohol and drug issues. Being realistic, the association will believe that the students will respond to the questions by giving socially desirable answers. Therefore, the organisation may hire a counsellor/researcher from a local University in order to meet the students, gradually build a strong rapport with the students, and this will result in the students feeling calm and comfortable around the counsellor/researcher, meaning they are considerably more likely to respond to questions with authentic, realistic and truthful answers. (Cochran, 1977) Planning prior to conducting a quota-sampling greatly minimises differences, ultimately leaving the researcher with more accurate and precise results. A significant advantage of using quota sampling is that it can be used when random sampling is impossible; quota sampling is also a very simple process that is quick to carry out, and therefore an ideal form of sampling when restricted by time constraints. However, within the quota sample, biases may still exist, with them being difficult to eliminate as they are not controlled by the quota sampling. (Kalton, 1983)
Stratified sampling involves putting the members of the population into categories/groups. The advantages of using stratified sampling are that is focuses on the priority subpopulations, ignoring the less relevant subpopulations. Stratified sampling also allows the use of different sampling techniques for different subpopulations, this considerably improves the overall accuracy of the hypotheses and result, in addition to being a practical and valuable solution to sampling when the population is too large to use in one long list. However, the selection of relevant stratification of variables is difficult to achieve, with the data not permanently useful when there are no identical or similar categories/groups. Stratification sampling is also an expensive form of sampling as it requires accurate information about the population that is being used, with the risk of biases being introduced due to there being errors within the measurements, or a clear bias when the selection takes place. Stratified sampling can also be combined with other sampling techniques in order to achieve the most accurate results possible. (Barnett, 1991)
Quota sampling and stratified have some very clear similarities. Both specify the number of subjects that are to be included within the sample based on selected characteristics. The function of quota sampling is to ensure the sample gives an accurate, genuine and realistic representation of the population regarding important characteristics that have been put into place prior to the sample being conducted. This is achieved through subjects with specific characteristics that have been placed into sub-categories containing similar populations. Stratified sampling can be seen to take place in order to ensure adequate and ample numbers for sub-analysis once the sample has been concluded. (Foreman, 1991)
Many respected organisations and researchers are forced to rely and use non-random sampling due to random sampling being difficult to accomplish. Non-random sampling can be clearly justified if it is highly unlikely, or impossible, to conduct a truly random sample. As a result of these difficulties, the organisations and researchers are most likely to resort to conducting a quota sample or stratified sample. (Kalton, 1983)
Cluster sampling involves the population being divided into groups, or clusters. The researchers involved randomly select the clusters to be included in the sample, with each element being assigned to one group solely. Providing the size of the sample is continuous across all the sampling methods, cluster sampling does not provide as much accuracy as other sampling methods, namely random sampling and stratified sampling. Thus, it is logical for people to ask, when conducting a sample, why use cluster sampling? With the answer being when using a limited budget to conduct the sample, the researcher(s) will be able to use a bigger sample using the cluster sampling method, with the increased size of the sample compensating for and counteracting the deficit of precision. Therefore, when on a limited budget for a sample, cluster sampling may be seen as the most appropriate and suitable method to use. (Jessen, 1978)
Convenience samples are conducted through the researcher, at their own convenience and discretion, choosing whether to make a valid attempt to ensure the sample is an accurate representation of the general population. An example of this is the researcher standing in a public area, for instance a shopping centre, and choosing who to stop and ask question and/or fill out a survey. Therefore convenience sampling is a form of non-random sampling, meaning the data obtained is inconsistent and does not give an accurate representation of the whole population. Whilst it is difficult to generalise the results of a convenience sample, they can still be informative, although not used by many to gain information and statistics. (Wiley, 1992)
When choosing a sampling method, it is essential to choose the method that most effectively links the particular goals involved. Budget can greatly affect the sampling method chosen, therefore it is important to ensure the sample is as accurate and precise as possible, identifying the types of sampling method that will bring the best results, achieving the goals set prior to sample being conducted. Therefore, it is not possible to choose an outstanding sampling method, as each method is dependant on a variety of factor, as mentioned before such as budget, population size and time constraints. It is the responsibility of the researcher to pick the best method, studied to the sample in order to illustrate a fair and realistic representation of the population.
Bibliography
Barnett, V. 1991. Sample Survey Principles and Methods. Edward Arnold, London, 173pp.
Cochran, W.G. 1977. Sampling Techniques, third edition. John Wiley & Sons, Inc., New York, 428pp.
Foreman, E.K. 1991. Survey Sampling Principles. Marcel Dekker, Inc., New York, NY.
Jessen, R.J. 1978. Statistical Survey Techniques. John Wiley & Sons, Inc., New York.
Kalton, G. 1983. Introduction to Survey Sampling. Quantitative Applications in the Social Sciences 35, Sage Publications, Beverly Hills, CA, 96pp.
Levy, P.S. and S. Lemeshow. 1991. Sampling of Populations: Methods and Applications. John Wiley & Sons, Inc., New York, 420pp.
Thompson, S.K. 1992. Sampling. John Wiley & Sons, Inc., New York, 343pp.
