Statistical Methods in Human Resource Research: Dearth or Diversity? Evidence from Published Literature

Abstract

This article attempts to find out if there is breadth in application of quantitative techniques in published literature within the field of human resource management (HRM). In addition, it investigates the holistic use of specific categories of statistics, and if there are categories that are neglected. The study utilises a combination of research questions and hypotheses. The broad categories of statistics that this study focussed on include descriptive, data science statistics, exploratory graphical, advanced statistics such as structural equation modelling, Bayesian statistics and inferential statistics. It goes further to study application of machine learning statistics in HRM research. Using archival methodology, the article utilises a sample of 120 journal papers to answer formulated research questions and hypotheses. Descriptive statistics, exploratory graphical analysis and inferential statistics are used in the analysis. The findings indicate that there are neglected statistics in HRM research. Overall, most statistical categories are underutilised. HRM journal editors, researchers and practitioners must stock HRM methodological toolbox.

Keywords

HRM statistics Data Science HR Research E HRM

Introduction

There seems to be dearth of studies investigating statistical methods used in human resource management (HRM) studies. One of a small population of these studies was conducted by Laken et al. (2018) who argued that HRM has traditionally relied on general linear modelling which does not always fit the hierarchical and longitudinal data collected by HR function. Additionally, Shen (2015) notes that HRM field has lagged behind other fields of study as far as adopting multilevel modelling is concerned. The researchers went on to propose the use of bathtub modelling and optimal matching analysis. Medical field has robust and diversity of studies (Arnald et al., 2013; Nieminen et al., 2018) investigating the statistical models used in this field. Information and library science has a good number of studies (Zhang et al., 2016, 2018) exploring the use of statistics in the field. Statistical analyses can broadly be broken into two categories: confirmatory (CDA) and exploratory analysis (EDA). Nicebmboim et al. (2018) advice that when the two are used together, clear separation should be maintained so as to allow the researcher to check whether results from exploratory stage are robust. Adding their voice, Kimmelman et al. (2014) point out that researchers should disentangle the exploratory and confirmatory statistics while utilising their complimentary nature. They argued that each uses different study designs and supports different inferences. In addition, and according to the authors, the separation makes replication of studies tenable. Gaus et al. (2015) explain that researchers must indicate if statistical tests are confirmatory or exploratory. Salkind (2010) argues that EDA was not conceived as a substitute for CDA, but its application serves a different stage of the research process. Komorowsi et al. (2016) outline the use EDA as being used to examine a distribution, outliers and anomalies. They add that EDA can also be used in hypotheses generation unlike CDA which is used in hypotheses testing. Despite the fact that hypotheses testing predominates quantitative research, there is a reservoir of value in hypotheses generation through exploratory statistical analysis. Though most quantitative researchers are obsessed with theory and model testing, quantitative analysis can be used as effective pathway in theory and model building. Roberts and Grover (2009) explain structural equation modelling (SEM) a quantitative technique as having a significant ability for specifying, accessing and modifying theoretical models. Tarka (2018) discusses that the growing need to utilise SEM was as a result of researchers wanting to understand the structure and interaction of latent phenomena. According to Sreiner (2006) SEM works by incorporating both confirmatory factor analysis (CFA) and path analysis. Social sciences research has been accused with using less qualitative analysis and more of quantitative analysis. We can apply an inward introspection on quantitative analysis as contained in published literature so as to discern if there is prevalent use of select statistical analysis and neglect of other quantitative methods. Even within particular broad categorisation of statistics, the researcher will dissect further to find out whether the breadth of various specific categories is utilised. Exploratory data analysis made popular by Tukey (1977) book, has a lot to offer in quantitative data analysis. Data science analysis techniques such as, neural networks are gaining popularity in social science. There is a close relationship between data science and statistics. Blei and Smyth (2017) discuss data science from three perspectives: statistics, computer and human. There are various studies published of late modelling HRM phenomena by using neural networks (Chandraseker et al., 2015; Margarita et al., 2018; Perez-Campdesunar et al., 2018; Somers et al., 2018; Wang & Shun, 2016) and other associated statistics. For research to give holistic insights and patterns, then the breadth of statistics must be applied so as to tease out hidden patterns.

Research Questions

This research will be guided by the following research questions:

How do journal papers utilise breadth of quantitative methods?

How do journal papers utilise breadth of specific statistics within broad categorisation of quantitative methods?

Hypotheses

The following two hypotheses will be used in this research:

H₁: There is no significant association between statistical category and journal publisher.

H₂: There is no significant association between statistical category and application area.

Literature Review

This article discusses several statistics that are usually applied in HRM research and various categories. It will explore the usage of these broad categories as well as specific ones under each category. There are broad categorisation of statistics that were considered in this article. The article will also categorise the statistics based on parametric and non-parametric categories. Simpson (2015) advices that the first step in data analysis is to describe data collected using figures that give visual impression of data. In addition, the author advices that researchers should then use numerical descriptions using statistics. Inferential statistics should then be applied. Arnald et al. (2013) did a research in medical field and reported that there was increased use of advanced statistics such as multivariable regression, multilevel modelling, survival and sensitivity analysis. In another research within the same field, Nieminen et al. (2018) found that traditional statistics for comparing independent groups such as chi-square test, t-test and ANOVA were commonly used. Within library and information science Zang et al. (2016) found a growth trend in five out of six journals in the usage of statistics. They reported t-test, chi-square, ANOVA as the most frequently used statistics. In a related study the same authors Zhang et al. (2018) found significant relationship between statistical methods and application areas of studies performed. Kaur and Fink (2016) argue that advanced technologies such as machine learning and artificial intelligence are at experimental stage in HRM. They add that ‘talent analytics’ is also called ‘people research and analytics’, ‘workforce analytics’, ‘HR analytics’, ‘business HR analytics’ and ‘HR business analytics’. Grillo and Hacketti (2015) note that HR analytics is advancing from descriptive to prescriptive and predictive analytics. Jebb and Woo (2017) discuss exploratory data analysis as a mode of analysis that deals with exploration, discovery and empirically determining a phenomenon.

Methodology

The study utilises published papers contained in reputable publishers in the field of HRM. The study utilised positivist philosophy. Two phrases were used in searching candidate papers for analysis. The analysis concentrated on two areas of HRM and thus two phrases were used that include electronic HRM (E-HRM), and HRM and performance. Papers that were displayed after search criteria were downloaded for further analysis. Only papers utilising quantitative paradigm were selected. A sample of 120 papers was utilised from two journal publishers that included SAGE journals and Emerald across many journal titles. Data was analysed quantitatively using exploratory and descriptive statistics, while the research adopted archival methodology. Data was collected using document review guide. Statistical Package for Social Sciences was used to analyse data. Graphical exploratory data analysis was first used in analysis, followed by descriptive statistics. Non-parametric test was applied to see if there are differences between two journal publishers and the two areas of HRM in terms of statistics used.

Findings and Discussions

From the dataset, the researcher made several general observations that most papers do not report statistics utilised, in the abstract only two papers did so. Additionally, in a considerable number of papers, the specific type of statistics is not mentioned, only the analysis method is mentioned.

Response rates are also not mentioned, only one paper mentioned response rate out of 120 papers. Table 1 indicates that most papers (85%) utilise parametric statistics while a few (15%) of them use a combination of both parametric and non-parametric statistics. No paper was reported to use non-parametric statistics exclusively. From Table 2, descriptive statistics exclusively are the most used (25%) in analysing HRM data. The same table indicates that overall, all statistical categories were poorly represented as none was used in more than 30% of papers sampled. A number of statistics categories were poorly used in sampled papers. These include exclusive use of exploratory graphical statistics (2.5%) and inferential statistics (3.3%). It was also found that use of combination of statistical categories was also poor apart from combination of descriptive and advanced statistics (30%). Advanced and descriptive statistics used together was second most used (9.2%) combination of statistical categories used to analyse HRM papers. Exploratory used together with descriptive was also poorly represented at 4%. Other combination categories that were poorly used include descriptive and data science combination (1.7%), descriptive, advanced and inferential (2.5%) and exploratory, descriptive, inferential and advanced (1.7%). Table 3 indicates that half of the papers were from SAGE journal publishers while the other half were from Emerald journal publishers. Table 4 indicates that 50% of the papers focussed on E-HRM as an application area while the other 50% focussed on HRM and performance as an application area. There is no diversity as far as use of regression is concerned in HRM papers as only a few categories registered some papers that utilised particular type of regression as shown in Table 5. The same table indicated that linear regression was mostly used (17.5%) type of regression to analyse data followed by hierarchical and moderated at 3.3% each. Logit regression was the last at 2.5%. Thus, HRM relies heavily on linear models of data analysis. This finding is consistent with opinion of Laken et al. (2018). Generally, only 26% of papers used regression analysis. Ridge, lasso multinomial, binary and probit regression methods were never used in the papers sampled. Moderated and hierarchical were coded because they appeared in a number of papers. Notably, both indicate wrong use of regression terminology. There was also lack of diversity in the usage of correlation as a method of data analysis. Table 6 indicates that Pearson correlation was mostly used (7.5%). As shown in the same table, 21.7% of papers that used correlation method never indicated which type of correlation was used. Only 30% of papers out of 120 used correlation as a method of data analysis. Bayesian statistics are neglected in HRM research as out of 120 papers, no single paper used Bayesian statistics. HRM researchers rarely use exploratory data analysis, despite the value that this category of statistics can add to HRM research. Exploratory graphical was only used to present data as shown in Table 7. Where exploratory was used alone or in combination with other categories of statistics as conceptualised in this article recorded less than 5% of papers as shown in Table 2. This indicates a glaring neglect of exploratory statistics in general within HRM research papers. Table 8 indicates that central tendency and dispersion combination were mostly (41%) descriptive statistics followed by frequency tables (18%) then followed by combination of central tendency, dispersion and frequency tables (10%). Table 9 indicates that only chi-square was used as non-parametric statistics. Even then it was used mostly together with SEM. Thus, there is generally very low use of non-parametric statistics in HRM research papers. As shown in Table 11 data science statistics was poorly used. When used with inferential statistics, there were 0.8%of papers that utilised this combination while 1.7% was recorded on data science combined with descriptive statistics as indicated in Table 2. From Table 12, combination of descriptive and inferential was mostly used (30 papers) in both E-HRM and HRM performance application areas followed by descriptive statistics used exclusively (10 papers). This indicated most of HRM papers (66 papers) used either descriptive statistics exclusively or a combination of descriptive and inferential statistics. Table 12 indicates that there is low use of combination of statistical methods in HRM research as most combinations of categories of statistics used posted less than 10 papers. This fact is collaborated by Figure 1. Table 13 indicates that across the two journal publishers, a combination of descriptive and inferential statistics (36 papers) and descriptive statistics used exclusively (30 papers) were the most used. As shown in Figure 2, generally there was marginal difference in usage of statistics or combination of statistics across journal publishers. However, Figure 2 indicates that data science statistics was very lowly used in papers published by both journal publishers. In Figure 3, advanced statistics was used more in HRM and performance application area than E-HRM application areas. Advanced statistics combined with inferential statistics was also used more in HRM and performance application area than E-HRM application area. Additionally, papers published on both HRM and performance and E-HRM application areas rarely used data science statistics. Figure 3 indicates that any combination of statistical categories that included exploratory graphical statistics recorded less than five papers indicating very poor utilisation of exploratory statistics. HRM research articles have neglected the use of non-parametric statistics as indicated in Table 9. Only chi-square as a parametric statistic was used sparingly but even then, was used as a test statistic for SEM. Apart from SEM, other advanced statistics such as multilevel modelling and meta-analysis statistics are not used in HRM research. Only one paper out of 120 was recorded for multilevel modelling and no paper was recorded for meta-analysis as shown in Table 10. These findings agree with Shen (2015), who pointed out that HRM research lags behind in multilevel modelling. Research hypotheses were tested using chi-square and results are contained in Tables 14 and 15. Table 14, indicates that there is no significant association between statistical category and journal publisher (X² (11) 13.318, P (.273 > .05) while Table 15 indicates that there is a significant association between statistical category and application area (X² (11) 23.967, P (.013 < .05).

Table 1.

Broad Category Statistics

Broad Category
		Frequency	Percent	Valid Percent	Cumulative Percent
Valid	Parametric	102	85.0	85.0	85.0
	Parametric and Non Parametric	18	15.0	15.0	100.0
	Total	120	100.0	100.0

Source: The author.

Table 2.

Statistics Category

Statistics Category
		Frequency	Percent	Valid Percent	Cumulative Percent
	Exploratory graphical	3	2.5	2.6	2.6
	Inferential	4	3.3	3.4	6.0
	Advanced statistics	9	7.5	7.8	13.8
	Descriptive statistics	30	25.0	25.9	39.7
	Descriptive, advanced and inferential	3	2.5	2.6	42.2
	Inferential and advanced	10	8.3	8.6	50.9
	Exploratory, inferential, advanced, descriptive	2	1.7	1.7	52.6
	descriptive and inferential	36	30.0	31.0	83.6
	Exploratory, descriptive	5	4.2	4.3	87.9
	Data science and inferential	1	.8	.9	88.8
	Advanced and descriptive	11	9.2	9.5	98.3
	Descriptive and data science	2	1.7	1.7	100.0
	Total	116	96.7	100.0
Missing	System	4	3.3
Total		120	100.0

Source: The author.

Table 3.

Journal Publisher Statistics

Journal Publisher
	Frequency	Percent	Valid Percent	Cumulative Percent
SAGE	60	50.0	50.0	50.0
Emerald	60	50.0	50.0	100.0
Total	120	100.0	100.0

Source: The author.

Table 4.

Application Area Statistics

Application Area
	Frequency	Percent	Valid Percent	Cumulative Percent
E-HRM	60	50.0	50.0	50.0
HRM And Performance	60	50.0	50.0	100.0
Total	120	100.0	100.0

Source: The author.

Table 5.

Type of Regression Statistics

Type Regression
		Frequency	Percent	Valid Percent	Cumulative Percent
	Linear regression	21	17.5	65.6	65.6
	Logit	3	2.5	9.4	75.0
	Hierarchical	4	3.3	12.5	87.5
	Moderated	4	3.3	12.5	100.0
	Total	32	26.7	100.0
Missing	System	88	73.3
Total		120	100.0

Source: The author.

Table 6.

Type of Correlation Statistics

Type of Correlation
		Frequency	Percent	Valid Percent	Cumulative Percent
	Unnamed_Correlation_Type	26	21.7	72.2	72.2
	Pearsons	9	7.5	25.0	97.2
	Canonical	1	.8	2.8	100.0
	Total	36	30.0	100.0
Missing	System	84	70.0
Total		120	100.0

Source: The author.

Table 7.

Exploratory Graphical Statistics

Exploratory Graphical
		Frequency	Percent	Valid Percent	Cumulative Percent
Valid	Just Presenting data	12	10.0	92.3	92.3
	Check data distribution	1	.8	7.7	100.0
	Total	13	10.8	100.0
Missing	System	107	89.2
Total		120	100.0

Source: The author.

Table 8.

Descriptive Statistics

Descriptive Statistics
		Frequency	Percent	Valid Percent	Cumulative Percent
Valid	Dispersion	1	.8	1.1	1.1
	Tables	22	18.3	23.7	24.7
	Central tendency and dispersion	50	41.7	53.8	78.5
	Dispersion, central tendency and tables	13	10.8	14.0	92.5
	Percentage	7	5.8	7.5	100.0
	Total	93	77.5	100.0
Missing	System	27	22.5
Total		120	100.0

Source: The author.

Table 9.

Non-Parametric Statistics

Non-Parametric Statistics
		Frequency	Percent	Valid Percent	Cumulative Percent
Valid	Chi square	22	18.3	100.0	100.0
Missing	System	98	81.7
Total		120	100.0

Source: The author.

Table 10.

Advanced Statistics

Advanced Statistics
		Frequency	Percent	Valid Percent	Cumulative Percent
Valid	SEM	33	27.5	97.1	97.1
	Multilevel modelling	1	.8	2.9	100.0
	Total	34	28.3	100.0
Missing	System	86	71.7
Total		120	100.0

Source: The author.

Table 11.

Data Science Statistics

Data Science Statistics
		Frequency	Percent	Valid Percent	Cumulative Percent
	Clustering_K Means	1	.8	50.0	50.0
	Neural networks	1	.8	50.0	100.0
	Total	2	1.7	100.0
Missing	System	118	98.3
Total		120	100.0

Source: The author.

Table 12.

Statistics Category and Application Area Cross Tabulation

Statistics Category *Application Area Cross Tabulation
Count
		Application Area		Total
		E-HRM	HRM and Performance	Total
Statistics Category	Exploratory graphical	2	1	3
	Inferential	2	2	4
	Advanced statistics	2	7	9
	Descriptive statistics	19	11	30
	Descriptive, advanced and inferential	1	2	3
	Inferential and advanced	2	8	10
	Exploratory, inferential, advanced, descriptive	2	0	2
	Descriptive and inferential	16	20	36
	Exploratory, advanced	5	0	5
	Data science and inferential	1	0	1
	Advanced and descriptive	2	9	11
	Descriptive and data science	2	0	2
Total		56	60	116

Source: The author.

Table 13.

Statistics Category and Journal Publisher Cross Tabulation

Statistics Category*	Journal Publisher		Cross Tabulation
Count
		Journal Publisher		Total
		SAGE	Emerald	Total
Statistics Category	Exploratory graphical	1	2	3
	Inferential	3	1	4
	Advanced statistics	4	5	9
	Descriptive statistics	15	15	30
	Descriptive, advanced and inferential	2	1	3
	Inferential and advanced	4	6	10
	Exploratory, inferential, advanced, descriptive	2	0	2
	Descriptive and inferential	16	20	36
	Exploratory, descriptive	5	0	5
	Data science and inferential	1	0	1
	Advanced and descriptive	7	4	11
	descriptive and data science	0	2	2
Total		60	56	116

Source: The author.

Figure 1.

Source: The author.

Figure 2.

Source: The author.

Figure 3.

Source: The author.

Table 14.

Statistical Category and Journal Publisher

Chi-Square Tests
	Value	df	Asymptotic Significance (2-sided)
Pearson Chi-Square	13.318^a	11	.273
Likelihood Ratio	17.239	11	.101
Linear-by-Linear Association	.181	1	.670
N of Valid Cases	116
a. 17 cells (70.8%) have expected count less than 5. The minimum expected count is .48.

Source: The author.

Table 15.

Statistical Category and Application Area

Chi-Square Tests
	Value	df	Asymptotic Significance (2-sided)
Pearson Chi-Square	23.967^a	11	.013
Likelihood Ratio	28.624	11	.003
Linear-by-Linear Association	.181	1	.670
N of Valid Cases	116
a. 17 cells (70.8%) have expected count less than 5. The minimum expected count is .48.

Source: The author.

Recommendations

Based on the findings, the following are recommendations to journal publisher’s fraternity in the field of HRM, editors of HRM journals as well as scholars and practitioners in the field of HRM. On a more general recommendation HRM papers should indicate response rate statistics and mention statistical methods used in data analysis in the abstract section. HRM researchers need to broaden their statistical toolbox. There is a need to broaden the statistics used as there is overreliance on a few categories. HRM researchers and practitioners must get skills in data visualisation by using of tables, graphs and charts. HRM researchers must increase their usage of regression as a method of data analysis. Additionally, they should utilise other regression methods such as ridge, lasso multinomial, logit, binary and probit regression methods. Researchers using regression in mediation and hierarchical fashions should use the right terminology, by indicating specific type of regression. Instead of using terms such as mediated and hierarchical regression, they may refer to them as an example, mediated linear regression and hierarchical linear regression. Researchers should use diverse types of correlations such as Spearman’s rank correlations, phi, canonical and Kendall’s Tau. HRM researchers should also increase the use of correlation as a method of data analysis. When correlation is used in HRM research papers, researchers should indicate the type of correlation that was utilised. There is a need to increase the usage of exploratory statistics in HRM research papers not just in presenting data but in determining statistics to be used in data analysis, as a precursor to CDA and for hypotheses generation. HRM papers need to utilise the power of data science statistics so as to arrive at interesting results based on the unique nature of these statistics. Owing to the large number of papers using descriptive statistics exclusively, there is a need to ensure that HRM papers utilise other statistics alongside descriptive statistics. It is recommended that HRM papers use combinations of exploratory, descriptive, inferential, Bayesian and data science statistics in same paper where necessary. Both journal publishers should endeavour to increase and champion utilisation of data science statistics and advanced statistics such structured Equation Modelling (SEM) statistics, multilevel modelling statistics and meta-analysis statistics in HRM research papers. Though SEM is being used in HRM as one of the advanced methods in data analysis, there is a high need to incorporate other advanced methods such as multilevel modelling (Gooderham et al., 2018; Laken et al., 2018; Shen, 2015) multilevel SEM, mixed modelling and meta-analysis. Researchers investigating EHRM application area need to incorporate advanced statistics such as SEM statistics, multilevel modelling statistics and meta-analysis statistics. When using descriptive statistics, researchers should strive to use the breadth of these statistics together including measures of central tendency, dispersion and frequency tables. HRM need to adopt the utilisation of Bayesian statistics in data analysis. Since there was no association between choice of statistics and journal publisher, the general recommendation would be for the two journal publishers to improve on the usage of diverse types of statistics in paper analyses. Researchers doing research on HRM application areas such as E-HRM and HRM and performance should strive to use breadth of statistics in data analyses.

Footnotes

Declaration of Conflicting Interests

The author declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.

Funding

The author received no financial support for the research, authorship and/or publication of this article.

References

Arnald

L. D.

, Braganza

M. S. R.

, & Coldits

G. A.

(2013). Statistical trends in journal of American medical association and implications for training across continuum of medical education. PloS One, 8(10), e77301.

Blei

D. M.

, & Smyth

(2017). Science and data science. PNAS, 114(33), 8689–8692.

Chandraseker

, Chidambaram

, Venkatraman

, & Venugopal

(2015). The viability of neural networks for modelling the impact of individual job satisfiers on work commitment in Indian manufacturing unit. Business: Theory and Practice, 16(3), 326–333.

Gaus

, Mayer

, & Muche

(2015). Interpretation of statistical significance-exploratory vs confirmatory testing in clinical trials, epidemiological studies, meta-analyses and toxicological screening. Journal of Clinical and Experimental Pharmacology, 5(4), 2–6.

Gooderham

, Fenton-O’ Creevy

, Croucher

, & Brooks

(2018). A multilevel analysis of the use of individual pay for performance systems. Journal of Management, 44(4), 1479–1504.

Grillo

M. A.

, & Hackett

(2015). What types of data analytics are being used in talent management organizations? Retrieved on July 10, 2020 from Cornell University, ILR School site: http://digitalcommons.ilr.cornell.edu/student/74.

Jebb

A. T.

, & Woo

S. E.

(2017). Exploratory data analysis as the foundation of inductive research. Human Resource Management Review, 27(2), 265–276.

Kaur

, & Fink

A. A.

(2016). Trends and practices in talent analytics. Society for Human Resource Management and Society for industrial and organizational psychology.

Kimmelman

, Mogil

S. S.

, & Dinargl

(2014). Distinguishing between exploratory and confirmatory preclinical research will improve translation. ProS Biolology, 12(5), 1–4.

10.

Komorowsi

, Marshall

D. C.

, Salciccioli

J. D.

, & Crutain

(2016). Exploratory data analysis In MIT Critical Data (Ed.), Secondary analysis of electronic health records. Springer.

11.

Laken

P. V.

, Bakk

, Giagkoulas

, & Bongenaar

& Leeuwen

L. V.

(2018). Expanding methodological toolbox of HRM researchers: The added value of latent bathtub models and optimal matching analysis. Human Resource Management, 57, 751–760.

12.

Nicebmboim

, Vasisthth

, Engelmann

, & Suckow

(2018). Exploratory and confirmatory analysis in sentence processing: A case study of number interference in Germany. Cognitive Science, 42(4), 1075–1100.

13.

Nieminen

, Toljamo

, & Vahanikkila

(2018). Reporting data analysis methods in high impact respiratory journals. ERJ Open Research, 4, 1–8.

14.

Perez-Campdesunar

, & De-Miguel-Guzman

, Sanchez-Rodriguez

, Martinez-Viva

(2018). Exploring neural networks in the analysis of variables that affect the employee turnover in the organization. International Journal of Engineering Business Management, 10, 1–11.

15.

Roberts

, & Grover

(2009). Theory development in information systems research using structural equation modelling: Evaluations and recommendations. In Dwivedi

, Lal

, Williams

, Schneberger

, & Wade

. (Eds.), A handbook of research on contemporary theoretical models in information systems (pp. 77–94). IGI Global.

16.

Salkind

(2010). Encyclopedia of research design. SAGE Publications.

17.

Shen

(2015). Principles and applications of multilevel modelling in human resource management research. Human Resource Management, 55(6), 951–965.

18.

Simpson

S. H.

(2015). Creating a data plan: What to consider when choosing a statistics for a study. The Canadian Journal of Hospital Pharmacy, 64(4), 311–317.

19.

Somers

M. J.

, Birnbaum

, & Casal

(2018). Supervisor support, control over work methods and employee wellbeing: New insights into non-linearity from artificial neural networks. International Journal of Human Resource Management, 1–23.

20.

Sreiner

D. L.

(2006). Building a better model: An introduction to structured equation modelling. Canadian Journal of Psychiatry, 51(5), 317–324.

21.

Tarka

(2018). An overview of structural equation modelling: Its beginning, historical development, usefulness and controversies in social sciences. Quantity and Quality, 52(1), 313–354.

22.

Wang

, & Shun

(2016). Applying back propagation neural networks in prediction of management associate work retention for small and medium enterprises. Universal Journal of Management, 4(5), 223–227.

23.

Zhang

, Zhao

, & Wang

(2016). A study on statistical methods used in six journals of library and information science. Online Information Review, 40(3), 416–434.

24.

Zhang

, Wang

, Zhao

, & Cai

(2018). Application of statistical methods in library and information science. Data and Information Management, 2(2), 103–120.