Factors To Consider When Choosing A Sample Size In An Essay.

Discuss About The Factors To Consider When Choosing A Sample Size?

One of the advantages of a large sample size is that it helps in reducing the margin of error hence improving the accuracy of the results from the sample. According to Cleary et al (2014), this will always make the population parameter to lie within the confidence interval of the point estimator. Being that the sample size chosen for the survey was 15,000, it was most likely to obtain more accurate information from the bank workers pertaining stress as compared to when smaller sample size would have been used by the two research institutions i.e. Katholike Universteit Leuven and the private company in Belgium. Additionally, large sample sizes are important in that they are representative of the wider range of elements contained in the population. Due to this therefore, all or most of the outliers will be captured in the sample unlike when the sample is size is small Belli et al (2014).

One major disadvantage of large sample is its costly nature. Reaching and covering a wider proportion of the population involves high expense incurred in the process Goodman et al (2013). The Union of Belgian Banks would therefore incur much through the assigned research institutes in the data collection from the targeted 15,000 bank workers. Being that the participants were not found from the same bank institution, the process would also be time consuming to get to various parts of the country for other bank workers.

Cost that will be involved in obtaining the sample is one among other factors that should be considered when coming up with which sample size to use in a survey. The risk involved in the values collected from the sample will also act as the determinant of the sample size i.e. if only the risk value matters in the collected values then as a result therefore, low risk values will call for large sample sizes. Prior information about the topic of study will also help in either reducing or increasing the sample size since prior estimates of means and variances will be used to help dealing with variation that could be found within groups (Button et al, 2013).

Sampling method is the process by which representative groups are selected from the population that is being studied. In the selection of the 15,000 bank workers for the sample by the research institutions, they used stratified sampling method. One of the advantages of the currently used sampling method by the institutions is that the sampling method minimizes sampling errors Ye et al (2013). This is achieved though dividing the population into subgroups called strata. The strata are spread to ensure that each characteristic of the population is represented by the strata and having the elements in each stratum selected by simple random sampling hence reducing sample selection bias. Also, stratified sampling method ensures that the targeted population is highly represented in the sample. Difficulty to identify ways of subdividing the population into subpopulations makes it at times unusable by the researchers, where this forms one of the major disadvantages. Additionally, stratified sampling method is time consuming where a lot of time is spent in the identification of the strata and then later select the sample from each strata through simple random sampling method (Acharya et al, 2013). In relation to the situation at hand, the research institutions i.e. the private specialized company in stress at work and Katholike Universteit Leuven, they first had to identify all the bank institutions in Belgium then divide the workers according to their bank institutions to form strata where further, the workers were now to be selected from the bank institutions through simple random sampling method to provide equal chances of obtaining the workers that will form the useable subset of the population. In order to improve the effectiveness of this sampling method (stratified sampling method), I therefore suggest that the number of strata to be increased. This will increase from where sampling of the individuals in the population will be selected hence representing almost the entire population thus reducing the marginal error in sampling.

Research Design

Cross-sectional design is a tool used by the researchers to obtain specific point time information from the collected data. It has some of the advantages and disadvantages. One of the advantages as noted by (Shen and Björk, 2015) is that the cross-sectional research design through cross-sectional study can help in ascertaining the worthiness of assumptions in the study. Also, as compared to other research designs, cross-sectional design is less time consuming. Since information is for specific point time of the already collected information, it therefore take cross-sectional research design less time to identify information of interest. Furthermore, the research design inexpensive. Unlike cross-sectional research design, longitudinal design has the potential to display the pattern of variable or variables for a certain period of time as it major advantage.

Disadvantages of cross-sectional design is that it cannot be relied on to predict the relationship between and the findings this is due to it lacking the time element since it only measures point time information. Prevalence as a result of extended period of time cases, these are seen from the cases that exist for a long period of time and they may be perceived less serious. On the other hand therefore, longitudinal research design is more expensive since it covers a long period of time. Longitudinal design is as well time consuming due to its ability to predict pattern over period of time. Also, when the expected outcomes are less, longitudinal design becomes less efficient (Shen and Björk, 2015).

In the collection of data from the bank workers, the research institutions used questionnaires that were structured with questions where the respondents were only required to give their responses on the provided spaces. Just like any other method of data collection methods, collection of data through questionnaires face some of the problems that need to be addressed.

According to Chernick et al (2011), respondents’ dishonesty is one of the major problems that questionnaires have. Respondents can decide not to be truthful when they are responding to the questions with the fear that their identities can be disclosed to the public. This can tamper with the accuracy and reliability of the results if such happens in the process of data collection. To eradicate this problem, the researcher is supposed to assure the participants who take part in the process that their privacy is highly valued and that they be kept private without access of any unauthorized persons. When this is effected, such problems would not reoccur or they will reduce in future.

Procedure of Data Collection

Being that the questionnaires were prepared by the research institutions and distributed to various banks in the country, there was no physical touch or face to face communication between the researchers and he respondents, the respondents will respond to the questions according to their own understanding of the questions and interpretations. For the same results intended by the subject of study, the questions if not clarified may not result to the common understanding. This can be as a result of unclear questions to the respondents. This problem therefore can be dealt with by the researcher through creating or composing questions that are easy and simple to understand and answer.

Difficulty of the questions to analyze is another problem that questionnaires have. When constructing the questions in the questionnaire, if the questionnaire happen to contain so many open ended questions, this will call for the opinion of the respondents hence cannot be coded during the analysis process. This occur more often whenever there are open ended questions and people’s opinions vary from one individual to another thus resulting to too much data that cannot be handled with ease and analyzed. This problem therefore can be dealt with and corrected by coming up with good question types that are close ended and that will allow for multiple choices that can be coded for easy analysis.

Face to face communication has been so effective in that one can be able see the emotions of a person through facial expressions, but with questionnaires it is difficult to capture such emotional responses that are expressed by the respondents especially when the questionnaire is administered thus data that could be observed from the respondents on their body language would be lost or go unnoticed. This particular type of problem can be combated by constructing questionnaires that have Likert scale that would be used to rate the attitude, feelings or emotions of the respondents.

At some times, the provided questions in the questionnaires are not always responded to. The respondents may decide to skip some of the questions due to their own reasons and submit the questionnaire forms with the skipped questions unanswered. For the case of online survey, they normally come with clear solution to such kind of problem. They simply make all the fields for the questions required without which the respondent cannot proceed to the next step or the form cannot be submitted. But for the case of our research question questionnaires, we can combat this problem by constructing uncomplicated questions and above all make the survey short, this will help increase the completion rates.

Accessibility to the questionnaires is another major problem that is faced when data is collected using questionnaires. Questionnaires do not always take care of people with some forms of disabilities such as visual or hearing impairment. Such people are not suitable to use the questionnaires and this can be corrected by using questionnaires whose accessibility options are built in.

Secondary data are second hand information that are obtained from the database archives. Depending on the subject of study, secondary data that will be used must be relevant to the subject of study. Also, before the secondary data is used to check for the representativeness of the sample, competency and accuracy of the data to the subject of the study must be first checked and confirmed, this is according to Piwowar and Vision (2013). Using secondary gives the researcher a clear picture of what he/ she expects and therefore saves time. Most of the secondary data are always obtained from the databases where they are stored making their retrieval easy and cheaper as compared to collecting primary data. Secondary data that were funded and collected by the government are in most cases involving large samples which result to the increased statistical precision since larger proportion of the population is represented.

Understanding secondary data can be done through reading the manuals that are stored alongside the data where thereafter they should be prepared for use in checking for representativeness. All variables and the treatment of missing data should be appropriately addressed to hold the meaning of data. Suitable sampling design mostly probabilistic sampling designs are supposed to be applied where since the sample is large, stratified sampling method is seen appropriate since it always represent more items from the population. Statistical analysis to be used is supposed to be ensured that it reflects the sampling design that was used where the point estimates such as means, variance and standard deviations should be in a manner that they cater for unequal sampling probabilities. The obtained secondary point estimators are then compared to the primary point estimators of the subject of study. If they are onto each other or too close to one another, then there will be confidence that the obtained point estimators are the reflection of the population parameter and thus the data is representative.

References

Acharya, A.S., Prakash, A., Saxena, P. and Nigam, A., 2013. Sampling: Why and how of it. Indian Journal of Medical Specialties, 4(2), pp.330-333.

Belli, S., Newman, A.B. and Ellis, R.S., 2014. Velocity dispersions and dynamical masses for a large sample of quiescent galaxies at z> 1: Improved measures of the growth in mass and size. The Astrophysical Journal, 783(2), p.117.

Button, K.S., Ioannidis, J.P., Mokrysz, C., Nosek, B.A., Flint, J., Robinson, E.S. and Munafò, M.R., 2013. Power failure: why small sample size undermines the reliability of neuroscience. Nature Reviews Neuroscience,14(5), pp.365-376.

Chernick, M.R., González-Manteiga, W., Crujeiras, R.M. and Barrios, E.B., 2011. Bootstrap methods. In International Encyclopedia of Statistical Science(pp. 169-174). Springer Berlin Heidelberg.

Cleary, M., Horsfall, J. and Hayter, M., 2014. Data collection and sampling in qualitative research: does size matter?. Journal of advanced nursing, 70(3), pp.473-475.

Goodman, J.K., Cryder, C.E. and Cheema, A., 2013. Data collection in a flat world: The strengths and weaknesses of Mechanical Turk samples. Journal of Behavioral Decision Making, 26(3), pp.213-224.

Kühberger, A., Fritz, A. and Scherndl, T., 2014. Publication bias in psychology: a diagnosis based on the correlation between effect size and sample size. PloS one, 9(9), p.e105825.

Piwowar, H.A. and Vision, T.J., 2013. Data reuse and the open data citation advantage. PeerJ, 1, p.e175.

Shen, C. and Björk, B.C., 2015. ‘Predatory’open access: a longitudinal study of article volumes and market characteristics. BMC medicine, 13(1), p.230.

Ye, Y., Wu, Q., Huang, J.Z., Ng, M.K. and Li, X., 2013. Stratified sampling for feature subspace selection in random forests for high dimensional data. Pattern Recognition, 46(3), pp.769-787.