Developing a Sampling Plan

Sampling Plan

Before discussing a sampling plan, there has to be clear and unambiguous definitions of what a sample and sampling are. Despite diversity in the definition of a sample, the best meaning is that a sample could be considered as a subset of a population, with which a researcher would like to use as participants in a given research study (Landreneau & Creek, 2012). According to Deming (1990), sapling is a science, which specifically guides quantitative studies, materials, behavior and the different causes of difference. In other aspects of research such as the qualitative research, sampling could be considered as the art of selecting a part of a population, in a given research area that is a representation of the entire population.

Both the qualitative and quantitative researchers approach their sampling differently. For the quantitative researchers, samples which are selected are those that will give the researcher easy time when trying to achieve statistical conclusions that are valid and also generalize their final results. Because of this reason, they have to come up with a sapling plan that is specific on matters of participant selection. For the qualitative researchers, they focus on sample plans that will develop rich and holistic understandings the phenomenon of interest (Polit & Beck, 2007).

Factors to be considered when developing a sample plan

Sampling frame

When developing a sampling plan, one of the most fundamental objectives is to come up with a favorable sampling frame. Basically, a sampling frame is the list that entails the population that the researcher intends to take selections from. A sampling frame has to be complete and current list of the population a researcher intends to research on. The best example would be the use of surveys in data collection, where the sampling frames are developed before the registration of the population. The best examples of a sampling frame would be telephone directories, attendance lists for schools and workplaces and voting registers (Denscombe, 2007).

Developing a sampling frame

When a researcher wants to device an efficient sampling frame, it is of great importance to actually distinguish between study populations and the target population. In many circumstances, the sampling frame is gotten from the study population, where there is an assumption that the study population reflects characteristics of a target population (Gerrish & Lacey, 2010). Especially for the researchers who are conducting surveys, they are obliged to search for the most suitable sampling frame. The researcher should, therefore, go to greater lengths of getting labels for purposes of posting, purchasing addresses from specific computer databases and many other sources (Denscombe, 2007).

Biasness in sampling frames

Researchers have to be aware, and ensure that there is no danger or possibility for biasness. In the first place, the researcher has to ensure that the sampling frame is not by any chance out of date. For instance, a private address research is not supposed to include those who are in the streets. Any electoral register is not supposed to be inclusive of those below the age of 18 years and are not ready to vote. If by any chance the sampling frame excludes items that have to be included, or includes things that are not supposed to be there, then the frame being developed will inevitably succumb to biasness. It is relevant that the researcher checks on the completeness of the sampling frame, to ensure it is up-to-date (Denscombe, 2007).

Elements of a good sampling frame

For a sampling frame to be considered efficient and effective, there are elements that have to be presented in the frame. These include; relevance, meaning its contents should be linked directly to the topic under research. It should also be complete in that all the relevant items needed in it have to be considered. Preciseness is the other factor requiring attention. The sampling plan has to exclude any items of irrelevance. Lastly, the sampling frame needs to be up-to-date. This means that any recent changes and additions have to be incorporated. On the other hand, redundant items have to be completely excluded from the frame unless they are necessarily required (Denscombe, 2007). The frame will be essential when identifying and selecting the best sample in a certain probability sample design, and also establishing contact with the selected elements (Sarndal, Swensson & Wretman, 2003).

Sampling units

According to Arturo and Gomez (1984), a sampling unit is the unit on which actual measurement is made. In case a given plot is the population of an area, then the sampling unit has to be smaller than the plot itself. An appropriate sampling unit has to be made and selected in order to attain the conditions of the research experiment. A sampling unit can otherwise be defined as an element, or many elements which are eligible for selection during the sampling process. For instance, a given airline could decide to choose every 20th passenger that is supposed to fly in a given date. Considering the example, the sampling unit will definitely be same as the element. On the other hand, the airline could decide to choose specific flights which would be the sampling units, then further choose passengers from the flights, who would be the elements of the sample (Zikmund & Babin, 2012).

Features of an appropriate sampling unit

An effective sampling unit has to consider some specified elements for appropriateness. One of the factors is the ease to measurement. When choosing a sample unit, the researcher has to consider that the unit will be measured easily to avoid future inconveniences. The unit also has to be of high precision and of low cost on the other hand. Precision could easily be measured using reciprocal of variance of a given sample unit. Cost, however, can be measured and based on the spending of time when measuring samples. If the sampling measurement time is short, then the costs are likely to be reduced. In order to keep the costs down and retain a reputable precision, then the researcher has to ensure the variability of the sampling unit (within a given plot) has to be small in size (Gomez & Arturo, 1984). For reference purposes, it is necessary to define what a primary sampling unit (PSU) is. When sampling surveys, PSU's arise when population elements are categorized and the different categories then become the units in the sampling process (Lepkowski, 2010).

Sample size

In most cases, the sample size is the most fundamental feature that affects any empirical study. Practically, the sample study is based on the expense of collecting data and also the required statistical power. When selecting a sample size, different methods could be applied in determining the sample size. These include; expedience, use of target variance and using a target with the statistical test. When determining the sample size, the precise approaches that need to be adopted include confidence intervals, computer intensive methods and power analysis (Dattalo, 2008).

Factors of consideration when determining sample size

Variables of measurement

The researcher has the task to decide on the appropriate variable to incorporate when calculating the sample size using formulas. For instance, if one chooses the seven- point scale when taking measurements of a continuous variable such as job satisfaction, and wants to determine whether the study respondents are different by specific categorical variables such as gender and educational levels, the best variable to use. If not, the primary variable (gender) will most likely result to a larger sample size than expected if seven-point scale was to be used (Bartlett, Kotrlik & Higgins, 2012).

Variance estimation

One of the critical components of formulas of sample size is when estimation of variance is a variable of study interest. Though the researchers have no control over the variance of the study, they have to integrate the variance estimates into the research design. There are different ways of estimating population and sample sizes. One of the ways to estimate is to divide the sample into two different steps, and then use the results acquired in the first step to determine the number of additional responses to consider in the next step. The second estimate is use of pilot study results. The researchers could also take advantage of studies that were taken earlier in circumstances where they studies are same or the population is similar. The last resort would be to guess or estimate a structure of a population, as long as there is evidenced logically by mathematical results (Bartlett, Kotrlik & Higgins).

Target population and accessible population

A target population is the collection of individuals, cases, items and objects that a researcher is interested in, and expects to take a sample from. For instance, in the case of a healthcare case, the target population could be patients between the age of 17 and 50 years with pains for at least duration of a year. Accessible population could be defined as the portion of a target population. The portion is accessible to the researcher for the reasons of adopting a specific study. A sample can only be taken from a study population that is accessible, but not from target populations (Sim & Wright, 2000).

Determining the need for target population

The activities of determining the need of a target population reflect the process of gathering, then analyzing and interpreting information concerning the target population. There is a need to make evaluative judgments concerning the scope, nature and extent of needs. The above process is also known as the needs assessment (Maher, 2012). When dealing with the individuals that were in the target population, population validity can be retrieved. Population validity is the ability to effectively generalize a study from specific individuals that was conducted to the larger population (Burke & Christensen, 2010).

Stratification

A stratified sampling category is a sapling for probability. The technique is evidenced when a researcher divides a certain population to form different subgroups known as strata. Later, the researcher will select the subjects randomly as per the subgroups in the right proportions. In relation to precision, stratification is used to reduce any variation in and within the subgroups and relatively increase the precision for population estimates (Husch, Beers & Kershaw, 2002).

Sampling strategies

Generally, methods of sampling are two in number. It is, therefore, upon the researcher to decide which sampling method is necessary and effective for research. The researcher will have a choice between a probability sample and a non-probability sample. The probability sample can also be referred to as the random sample whereas the non-probability sample can be known as the purposive sampling. The major difference between the two sampling categories is that probability samples are more prone to the fact that those selected from a wide population can easily be known whereas in non-probability samples the chances of a member being selected from the wider population is unknown. In the cases of probability samples, all the members of a target population could be chosen and they all have equal chances to be included in a sample. It is vice versa for non-probability sampling, where a few members of a wider population are definitely going to be considered, as a few are left out. The members do not have equal chance when dealing with the latter sample method (Cohen & Lawrence, 2007).

Probability samples

Due to the fact that the probability sample draws randomly from the wider population, the sample method will be relevant for researchers who are willing to make generalization. This is because the probability sample will always fight to seek representation of the wider population. In most cases, probability sampling is used in randomized controlled trials. In contrary, non-probability sampling generally avoids representations of the wider population and just represents specific groups and named sections of the wider population such as a class in a particular school or a number of students among examination candidates.

When considering the aspects of biasness, the probability sampling tends to have a limited rate of biasness when compared to non-probability sampling. This is ironical because it is the probability sampling that is supposed / expected to have biasness cases due to the fact that probability sampling covers a larger scope. This does not mean the probability sampling has no biasness, as this sampling method is prone to errors especially during the sampling process. This can be witnessed in opinion polls, where the researchers usually declare error factors such as positive or negative three percent. There are several types of probability samples including; systematic samples, simple random samples, stratified samples, cluster samples, multi-phase samples and stage samples. In all the mentioned samples, they have a common ground in randomness and therefore definitely have a degree of generalization (Cohen & Lawrence, 2002).

Systematic sampling

Systematic sampling is a type of probability sampling, and involves the selection of sample units in accordance with specified patterns of space and time. An example would be the selection of sample units with equal distances and time intervals. Systematic sampling can them be divided into two sub-sections, which are systematic grid sampling together with systematic random sampling. In the systematic grid sampling, the area of concern is subdivided using triangular grids or squares and the researcher later collects samples from the nodes and any specified locations. The first sample is randomly selected as the subsequent samples are gotten from the time intervals and fixed space. Contrary to the systematic grid sampling, there is the systematic random sampling where the area of concentration is subdivided into a number of grids, and a sample is collected from every grid created (Zhang, 2007).

The benefits that come with using systematic sampling are many. One is that using systematic sampling will assist in the extrapolation of information from the current periods to future periods using easier processes compared to other sampling methods. Also, seasonal cycles are easily identifiable and accounted for in the process of data application. The other benefit over the other sampling methods is the fact of the fixed sampling schedule, which will assist in administration of sampling. One other fundamental effect of the systematic sampling is that it makes the calculation of standard deviation and mean easy and straight forward. When analyzing the data, it will be simpler if the researcher assumes that the data was gotten through simple random sampling (Zhang, 2007).

Simple random sampling

Simple random sampling is assumed to be the basis of all the other sampling methods. When using the simple random sampling, each unit of a particular frame is given numbers from I to N (n usually represents the population size). After the first step, a table is developed that to assist in random number generation, which is computerized to give output and results for the random numbers. In most cases the numbers are random in different directions, and, therefore, any space in the table is there to ease the reading of values (Black, 2009).

Notably, the simple random sampling has limited application when it comes to large-scale surveys and large research investigations. Despite the limits it has, the simple random sampling techniques are simply designed theoretically and this allows for easier benchmarking. The simple nature of the sampling method allows for researchers to practically address more complex methods of sampling using the simple random sampling. In this sampling, cost efficiency factor of the design is measured by multiplying the design effect and the ration of cost per case regarding the design under consideration to the cost of every case (Marsden & Wright, 2010).

Simple random sampling is the most logical and fundamental technique among the probability sampling methods. Ironically, it is the least used in real practical researches (Maxfield & Babbie, 2011).The advantage of the simple random sampling is the fact that it is the easiest to implement, when compared to the rest. The other advantage could be that it is sufficient and very effective in case the researcher wants to provide for sub-sampling within the frame of the research. However, samples can contain extreme variable intervals between successive items that are being tested. The disadvantage with extreme variable intervals is that it may lead to poor choice when monitoring items in a continuous sequence (Ellison, Barwick & Farrant, 2009).

Stratified sampling

In stratified sampling, a given population is divided consciously into non-overlapping subgroups known as strata. After the division, a probability sample is then selected in each and every stratum. All the selections in the different strata are independent. Basically, the stratified sampling method is the most flexible and powerful technique which is applied by different researchers (Sarndal, Swensson & Wretman, 2003). The popularity of the sampling method directs its basis to a few factors explained in the next paragraph.

In case the estimated values of a specified precision are required for sub-populations, then stratified sampling will be used. Each sub-population is treated as a different stratum mostly when the membership of the domain is specified in the frame of the sampling plan. Also, in many surveys, aspects such as measurement, response and information that is auxiliary usually differ from sub-group to sub-group. This problem is dealt with by the stratified sampling techniques. The method also favors many administrative decision makers, who wish to divide territories into different geographical districts with representative offices in the districts. Basically, the managers usually let the district represent strata for easier differentiation (Sarndal, et al., 2003). When dealing with the stratified sampling method, the researcher ought to ask a number of questions that will assist in the efficient use of the method. These elements that need concentration include; the construction and development of strata and choosing the right sampling and estimation methods.

Cluster sampling

To begin with, the understanding of what clusters are will be relevant in this section. Cluster samples are those samples which are gathered in chunks and groups of different elements which are aggregates of certain elements in a given population. During the process of cluster sampling, clusters are gotten from the target group or population. After that, random samples of clusters are later selected from the population and elements are selected and integrated in the sample. Clusters are more heterogenic especially within groups and among groups. This is contrary to what is experienced in the stratified random sampling, where homogeneity is experienced within every group as heterogeneity reigns across groups (Sekaran & Bougie, 2010).

A determined type of the cluster sampling is the area sampling, where clusters consist of geographical areas for instance countries, specific boundaries and city blocks within specific localities. This sampling method can be used in different ways, for example sampling consumer needs prior to opening a convenience store in a certain place of a given town or city. Advertisements can also use the cluster sampling method especially when targeting local population and many other instances.

When compared to other methods of sampling, cluster sampling is advantageous in many ways. First, cluster sampling happens to be less expensive compared to other probability sampling designs and it is also not dependent on the sample frame. In fact, most of the researchers opting for the cluster sampling are motivated by the issue of cost reduction. On the other hand, cluster sampling is easily affected by the biasness and it is definitely the least capable for generalizing purposes. This is because many of those cluster samplings occurring in organizations hardly have any heterogeneous elements. In simple terms, conditions of intra-cluster heterogeneity and the inter-cluster homogeneity are never met (Sekaran & Bougie, 2010).

Cluster sampling is further categorized to multistage cluster sampling. The latter involves repeating two steps that is listing and sampling. Primary sampling units are first selected for instance churches or blocks and maybe stratified for purposes of sampling. Later on, a sample of the units is selected and vice versa. In summary, the multistage cluster sampling method brings impossible studies to possibility (Babbie, 2010).

Multi-phase sampling

Multi-phase sampling need not be confused with multi-stage sampling. In multi-phase sampling, the design used is such that information collected from the whole sample is combined with the information acquired from part of the previous/original sample. For instance, in case a survey was to be conducted aiming at knowing the circumstances affecting health facilitation and nature in a certain town, a detailed questionnaire could be used as a representation of the first phase of the survey, to find out the number of respondents that had used the hospital services and facilities of that particular city. The second questionnaire can then be given to those that testified having used the facilities. The second questionnaire will then be considered the second-phase of the research or study (Beri, 2007).

The main difference between the multi-phase and multi-stage sampling methods is that in multi-phase sampling, the differentiated observation phases relate to sample units of the same type. In the multistage sampling technique, sample units are different both in type and stage (El-Shaarawi & Piegorsch, 2001). One of the advantages of the multi-stage sample is that it is flexible. A researcher has the ability to utilize fully any available information of the respective and existing divisions and subdivisions of the entire population as the units according to different stages. Also, it is easier to conduct a multi-stage sampling process as compared to the single-stage sampling (Medhi, 1992).

Stage sampling

The two-stage sampling method, which is a simple version and replicate of multi-stage sampling, involves the random selection and sampling of clusters and later sampling members randomly from the selected clusters. The result is a final sample, in which the researcher can derive information from, conduct a survey, give questionnaires or run any other data collection technique. However, many researchers look for more flexible and simpler methods of sampling to avoid costs and acquire great results. Such flexible and efficient sampling methods include the simple random sampling (Adler & Clark, 2008).

Quantitative sampling

Choosing a sample of study is a fundamental step in the research project of any serious researcher. Sampling is important because it would be unwise and rarely practical to study huge populations due to the insufficiency it would cause. The main aim of quantitative sampling is to ensure the researcher draws clearly a representative sample from the entire population. This does not mean that the sample results affect the respondents only, as the results of the research are generalized at last, to represent the population (Marshall, 1996).

Qualitative and quantitative research is differentiated in numerous ways. This paper addresses just but a few. For quantitative research, the researchers have a desire to gat samples that will allow them to reach valid statistical conclusions and to later generalize their results as if they studied the whole population. The researchers, therefore, develop a sampling plan that will give details on advancement of participants and the number to be selected. On the other hand, qualitative research gives researchers an obligation or duty to develop a rich and established understanding of a specific phenomenon. The researchers are observant enough to decide on issues during the study time and they base their decisions of given theories and earlier conducted research. In this case, the researchers need not develop any sampling plan before they begin their studies and research (Polit & Beck, 2009). It is, therefore, of great relevance that a quantitative researcher understands the importance of studying a small sample because towards the end results reflect the entire population. This will happen only if the researcher chooses the right sample size, and provides a sampling plan that is effective in retrieving necessary information from the respondents of the research.

Quantitative sampling methods

The fundamental and key element, which assists many researchers to generalize in accuracy from a population sample to the whole population, is the probability sampling technique. Most of the probability sampling techniques, which are discussed in, detailed in the preceding paragraphs use random samples. Random sampling involves a precise and scientific procedure that is used in the selection of respondents to a given research. There is hardly any haphazard thing concerning the earlier stated statement. Specific sampling techniques are also adopted and used, in controlling and determining the likelihood that certain individuals will be selected as research respondents. For instance, when a coin is flipped before a game starts, so as to determine the beginners of the game, then both teams will have equal chances for selection. In such a case, the probability would be half for each team (50%). Other complex techniques also guarantee same probability in case there is selection of samples from the entire population (Rubin & Babbie, 2009).