Research And Statistical Method Management: Determining Sample Size And Importance Of Control Variables

Determining Appropriate Sample Size in Research

Discuss About The Research And Statistical Method Management.

The researchers with regards to their research face an issue in relation to determination of sample size which should be considered for the research. The key requirement of the sampling process is for the selected sample to represent the population of interest in a faithful and accurate manner.  It is noteworthy that in relation of determining appropriate sampling size, there is an inherent tradeoff involved which must be elaborated. In case of selection of a large sample size, the chances of the sample being representative of the population tends to increase but the downside is the additional costs involved in selection of this large sample and the associated cost of data collection.  On the other hand, with a lower sample size, it may so happen that the accuracy is compromised owing to the sample not being representative (Eriksson & Kovalainen, 2015).

Owing to the inherent trade off, it is essential to seek the determinants which enable the researcher in deciding the requisite sample size that would be necessary. One of these is obviously the accuracy which the researcher desires. In case of higher accuracy being desired, a higher sample size would be more desirable as compared to a situation where the accuracy requirements are low. Another factor is the heterogeneous nature of the population which would tend to determine the appropriate sample size. The mathematical expression for minimum sample size is also broadly based on these elements as highlighted below (Hair et. al., 2015).

Thus, in line with the above formula, it is apparent that higher the heterogeneity in the population of interest, higher would be the minimum sample requirement. Further, higher the MOE or Margin of Error that is acceptable to the researcher, lower would be the sample size requirement (Flick, 2015).

The current case needs analysis in the wake of above theoretical discussion. The population of interest for the given study consists of 69,000 employees belonging to Belgian banks and the selected sample size is 15,000 comprising about 21% of the population. The sample size seems appropriate taking into consideration the heterogeneous nature of population as there are employees from as much as 63 banks. Further, there are additional attributes which tend to be divided these employees such as levels, gender along with educational status. Any sample which is lower may lead to sampling errors and hence the chosen sample may not be representative of the population, thus compromising the results of the study (Hillier, 2016). In wake of the above discussion, it is fair that a large sample has been selected for the study under consideration.

Role of Control Variables in Ensuring Validity

The sampling method that has used to select the 21000 employees from the population of 63000 employees is simple random sampling. This is referred to a sampling method where all the elements comprising the population have an equal chance of getting selected. For the given research study, this may be carried out by labelling the employees with a unique integer. Then using computer program, 21000 numbers can be randomly selected from the pool of numbers allocated to employees. The employees corresponding to the numbers selected would form the sample for the given research. The advantages and disadvantages associated with the given sampling technique are highlighted below.

• The primary advantage of this sampling method is the convenience of use and limited knowledge requirement. This is quite necessary especially when the sample size chosen is significantly large (Flick, 2015).
• Since in this sampling method, there is no need of any classification, hence any errors which may be incurred on that count are absent in this sampling method (Hair et. al., 2015).
• It is an extensively used method which tends to result in representative sample selection especially if the size of the sample is sufficiently large considering the population characteristics (Eriksson & Kovalainen, 2015).

• The key issue with this sampling method is the biased sample which may be possible when the sample needs to accurate represent certain key attributes. This aspect can be substantiated with the example of the research study at hand. For the sample data, there are a lot of critical attributes like gender, employee level, bank name which have to be fairly represented. This tends to become difficult in case of random sampling since it can potentially happen that there is overrepresentation of a particular attribute and underrepresentation of another (Hillier, 2016). For instance, it if possible that a large proportion of male employees are selected while the proportion of female selected is comparatively small. Especially another concern is that the population data consists of employees of as many as 63 banks. As a result, there would be certain banks with limited employee strength. In simple random sampling, it is possible that such low representation to population may not be captured in the sample selected. Hence, it makes sense to first classify the population in accordance with the key attributes and then proceed with simple random sampling in each of these classifications.
• Considering higher biased, possibility potentially the standard error would be higher which would have an adverse impact on the accuracy and applicability of the results obtained (Flick, 2015).

For the given measures used in the research study, a relevant discussion in relation with the reliability and validity has been carried out as follows.

• Quantitative job insecurity – The measurement of reliability of this measure can be adjudged the use of cronbach’s alpha. The general rule is that for quantitative variables or measures, it should be above 0.8 for the reliability to be considered satisfactory. For the given measure, this parameter is in excess of 0.8 implying that reliability of measure does not pose any challenge.  Validity concerns also should not arise since in a similar context the measure has been used in a previous study. Hence, on both counts i.e. reliability and validity, there does not seem to be any concerns with the use of the given measure (Hair et. al., 2015).

• Qualitative job insecurity – The measurement of reliability of this measure can be adjudged the use of cronbach’s alpha. The general rule is that for qualitatative variables or measures, it should be above 0.7 for the reliability to be considered satisfactory. For the given measure, this parameter is in excess of 0.8 implying that reliability of measure does not pose any challenge.  Additionally, concerns on account of validity could potentially arise as only 10 out of the 17 measures have been considered and hence it may so happen that the researcher may have left out the crucial ones which potentially could undermine the validity of the results (Hastie, Tibshirani & Friedman, 2011).

• Psychological distress – Like the above measures, reliability for this measure seems fine considering a cronbach alpha in excess of 0.8. However, on account  of validity, there could be issues considering the fact that the questionnaire version used seems quite old and hence may not be valid in the current scenario. Over the last five decades, numerous studies have been conducted on psychological distress measurement and it would be preferred that a more recent measure in this regard is considered so as to pacify validity related concerns (Hillier, 2016).

On account of the above discussion relating to the key measures of various factors, it may be concluded that reliability does not seem to be an issue but the same cannot be concluded about validity which can be further improved.

The key objective of the given research study is to illustrate the level of association in the quantitative and qualitative measures related to job security and well-being. One of the interesting observations is that besides the above measures which are imperative for the given research, incremental variables such as education, age and gender have also been introduced in the research. These variables have been included as control variables as any modification in these inputs can lead to change in the dependent variable even though the independent variable may remain constant. Thus, changes in these variables can thereby adverse impact the result validity (Hastie, Tibshirani & Friedman, 2011)  Even though control variables are important, but it is noteworthy that these do not form the main concern of any researcher since the focus is on measuring values of dependent variable by altering independent variable. In order to let the researcher focus on the primary objective, the control variables are identified so that they can be kept constant for the study so that no effect on the dependent variable is on account of these control variables (Flick, 2015).

Through the example of the given research, the role of control variable can be explained. Take for instance the control variable age. It is likely that employees falling in higher age bracket would be more stressed and dissatisfied on account of quantitative measures considering that they are concerned at the prospects of searching a new job. Also, over the years, these employees would be expected to adapt to the quantitative issues which would cease to be that important. This is in sharp contrast with the employees that are younger in age. These tend to be ambitious and hence more demanding than their elder counterparts. Also, the quantitative measures of dissatisfaction would be less significant for these considering they are more adapting and hence can look for job elsewhere. For these employees, the qualitative factors are expected to be more critical (Hillier, 2016).

Also, similar to age, the other two control variables in the form of education level and gender would be critical too.  For those employees with higher education level, finding an alternative job may not be difficult owing to which the quantitative measures of dissatisfaction may not be too relevant. However, the qualitative measures are critical for these owing to the higher expectations that these education employees would have from their employer. In case of employers who are not very educated, retaining the job is the primary concern and the conditions are job are not critical. Hence, for these employees, the quantitative measures of dissatisfaction would be more significant and representative in comparison with qualitative measures. Similarly, owing to the gender roles, the prominent factors of dissatisfaction for both the genders would differ (Hair et. al., 2015). Considering the above, it makes sense to take these variables as control variable so that these cannot impact the dependent variable under study.

The relevant research design to be deployed for the given research is known as correlational research design. The associated positive and negatives of this research design are highlighted as follows.

• A key advantage is that in this research design, the amount of data collected is much higher in comparison with experimental studies. This may be related to the objective of the correlation design which is not to specify the precise relationship but to identify a trend between the variables of interest (Hastie, Tibshirani & Friedman, 2011)  
• Further this research design also serves as starting point for many other researches which tend to take the initial cue from such research studies and then use a descriptive research design in order to develop a causal relation which tend to show an association relationship (Hillier, 2016).
• Also, the correlation research design and the related results provide useful data and association relationship for other studies which can take these into consideration for the formulation of various hypotheses which can be further validated using experimental or descriptive design (Flick, 2015).

• A key limitation or disadvantage of using this research design is that it only focuses on commenting the strength and direction of association but not focus on understanding the underlying reason and the other factors which may be responsible for this association. Further, no causal relations can be derived from such research designs as these would require further validation (Hastie, Tibshirani & Friedman, 2011)  
• The use of this research design is appropriate when only two variables are involved as it cannot simultaneously account for more than two variables (Hair et. al., 2015).

References

Eriksson, P. & Kovalainen, A. (2015) Quantitative methods in business research (3^rd ed.). London: Sage Publications.

Flick, U. (2015) Introducing research methodology: A beginner’s guide to doing a research project (4^th ed.). New York: Sage Publications.

Hair, J. F., Wolfinbarger, M., Money, A. H., Samouel, P., & Page, M. J. (2015) Essentials of business research methods (2^nd ed.). New York: Routledge.

Hastie, T., Tibshirani, R. & Friedman, J. (2011) The Elements of Statistical Learning (4^th ed.). New York: Springer Publications.

Hillier, F. (2016) Introduction to Operations Research (6^th ed.). New York: McGraw Hill Publications.