In this study, quantitative approach is applied to observe the connection between each dimension and the outcomes of independent variables towards dependent variables. This chapter focuses on methodology with sections including conceptual framework, hypothesis, research procedure, data collection, measurement instrument, also data analysis methods, which will introduce the framework and hypothesis by presenting the correlations, make clear of the process when collecting data, and explain the reliability and validity of measurement instruments and methods to analyze data as well. The results of pilot test are presented to give an idea of further study and investigation, which were shown in Appendix B.
Research Framework
Figure 3.1 and 3.2 presented the framework, regarded as LTCKIP model, in this study.
Measurement instruments had explained the origins of the factors of each dimension, and most of the factors were adopted from former researchers’ study. The exception was transformational leadership, which was divided into five parts in multifactor leadership questionnaire (MLQ) after the proposal of Bass and Avolio in 1993 that claimed four factors.
Idealized influence was separated into two parts: idealized influence attributes (IIA) and idealized influence behavior (IIB). Idealized influence attributes represents the attribution of charisma because of the positive characteristics of leaders, for instance, power perceived, focusing on higher-order ideals and values, and so on. These make followers gain more trust and confidence, which build a strong emotional link to their leaders. Idealized influence behavior prefers to act upon the values which leaders emphasize that are usually collective sense of mission and aspects.
H4
Figure 3.1. Conceptualized LTCKIP model.
Figure 3.2. Simplified LTCKIP model.
Knowledge Sharing (KS) Technical Innovation (TI) Organizational Innovation (OI)
Research Hypotheses
According to the research questions, literature review and framework, the research null-hypotheses are conducted as follows:
H1: Transformational Leadership (TFL) has no effect on Trust (T).
H2: Transformational Leadership (TFL) has no effect on Organizational Culture (OC).
H3: Trust (T) has no effect on Knowledge Sharing (KS).
H4: Trust (T) has no effect on Innovation (I).
H5: Organizational Culture (OC) has no effect on Knowledge Sharing (KS).
H6: Organizational Culture (OC) has no effect on Innovation (I).
H7: Knowledge Sharing (KS) has no effect on Organizational Performance (OP).
H8: Innovation (I) has no effect on Organizational Performance (OP).
Research Procedure
Figure 3.3. Research Procedure.
Identify Research Questions and Hypotheses
Review of Literature
Develop Theoretical Framework Research Motivation
Develop Instruments Identify Problems
Develop Research Method
Translation and Expert Review of Instruments
Conduct Pilot Study
Review of Instrument
Proposal Meeting
Data Collection
Data Analysis Data Coding
Report Completion Conclusion and Suggestions
Final Defense
Thesis Submission Revision
The pace of this study is structured in subsequent process as Figure 3.3 shown above.
First, literature review and theoretical framework were achieved, and then research method and questionnaires were developed, which led to the following step of pilot study before thesis proposal. After proposal meeting for suggestions, data collection and analysis of main study were achieved, revision of former chapters, conclusion and suggestions for further study and research purposes were constructed into final thesis. Last, more revisions were finished after final defense of this thesis.
Data Collection
The target samples of the study were aimed to be employees in Taiwan. The respondents who stood in ICT and FinTech industry were mainly considered to be chosen, and requirements for the employees in order to ensure this study are listed below:
1. The working experience of the respondent in present company should be more than 6 months. In order to make sure the competence of respondents, it was considered that the standard of six months includes the time period of three-month probation period to go through and another three months to fully understand his or her work.
2. The respondents are working in Taiwan currently. Moreover, for the privacy and security of the respondents, confidential information of each individual was for academic use in this research only.
Sampling Method
In terms of sampling method, the study applied a non-experimental quantitative method to observe and implement the data. The synonym of non-experiment was regarded as correlation survey, which had less effect on samples and merely measured dependent variables of revelation to independent variable (Punch, 2014). Questionnaires were distributed online and by hard copies to respondents from ICT and FinTech industries.
Measurement Instrument
The questionnaires were collected in English version and translated by the researcher into Traditional Chinese version. By using quantitative approach, this study examined those responses to the correlation between each dimension in the conceptual framework. The instrument was composed of 6 variables with a total amount of 99 questions, which was examined in pilot test with possibility to be reduced according to the outcome. The questionnaire consisted of seven parts as follows as Table 3.1 shows: part 1) transformational leadership; part 2) trust; part 3) organizational culture; part 4) knowledge sharing; part 5) innovation; part 6) organizational performance; and part 7) demographics. For part 1 to part 6, the questions must be answered completely by the respondents, and each item rated with a 5-point Likert Scale ranging from 1) Strongly Disagree to 5) Strongly Agree. Respondents answered part 7 by selecting one from the multiple options for each question as well. The validity and reliability of each item were recognized in the authors’ original work suitably and achieved various validations from prior studies.
Part 1) transformational leadership, which was measured by multifactor leadership questionnaire (MLQ) and 20 questions were proposed by Bass and Avolio (1993); part 2) trust had 18 items from Shockley-Zalabak, Ellis, & Cesaria (2003); part 3) organizational culture owned 24 questions adopted from Cameron & Quinn (2006); part 4) knowledge sharing possessed questions adopted from Siemsen, Roth, & Balasubramanian (2008) and Sohail & Daud (2009); part 5) innovation had items from Van der Panne, Van Beers, &
Kleinknecht (2003); part 6) organizational performance owned questions from Yang, Marlow,
& Lu (2009). The validity and reliability of the entire items were tested in the pilot test.
Moreover, peer reviews and expert review were utilized to preserve the validity of the instrument as well. Pilot test was also conducted initially to ensure the validity of each item before gathering the complete data for the study.
Table 3.1.
Summary of Instrument Used in This Study
Part Variable Items Reference
1 Transformational Leadership (TFL) 20 Bass, Avolio, Jung, Berson (2003)
2 Trust (T) 18 Shockley-Zalabak, Ellis, &
Cesaria (2003)
3 Organizational Culture (OC) 24 Cameron & Quinn (2006) 4 Knowledge Sharing (KS) 11 Siemsen, Roth, &
Balasubramanian (2008);
Sohail & Daud (2009)
5 Innovation (I) 12 Van der Panne, Van Beers, &
Kleinknecht (2003)
6 Organizational Performance (OP) 9 Yang, Marlow, & Lu (2009)
7 Demographics 5 Designed by the author
Note. TFL= Transformational Leadership; T= Organizational Trust; OC= Organizational Culture; KS= Knowledge Sharing; I= Innovation; OP= Organizational Performance.
Reliability and Validity
Reliability
Reliability of instrument corresponds to the external and internal consistency of measurement, while validity of instrument represents the scales measured in order to prove and make clear of what researchers have evaluated essentially (Williams & Monge, 2001). In this study, Cronbach’s Alpha and Average variance extracted (AVE) were measured in this study to test reliability according to Table 3.2.
Table 3.2.
Reliability of Instrument Reliability
Type
Theoretical Meaning How Reliability was Achieved
Reliability Analysis
The steadiness or consistency of the measure of a variable (Neuman, 2011).
It is considered acceptable that Cronbach’s Alpha value being above .70 when calculating each variable (Nunnally, 1978).
Average variance extracted (AVE) is also taken into account. A cut-off value above or equal to .50 or higher is considered acceptable (Chin, 1998).
Validity
Four aspects for examining validity were presented in Table 3.3, not only the content of the questionnaires were adapted and tried to fit the respondents’ need to understand the questions, but Kaiser-Meyer-Olkin (KMO) and Bartlett’s test for Sphericity and factor analysis were measured as well.
Table 3.3.
Validity of Instrument
Validity Type Theoretical Meaning How Validity was Achieved Face Validity A type of validity
measurement, which an
The questionnaire is translated into Traditional Chinese by the researcher.
Content
Each item of the questionnaire refers to the definition of its variable, and is examined by two peer and expert reviews as well.
The items of questionnaires are adopted from previous validated measures, and had been undergone pilot test initially.
(Continued)
Table 3.3. (Continued)
Validity Type Theoretical Meaning How Validity was Achieved Construct
converge or how well the indicators of different construct diverge (Neuman, 2011).
Construct validity in the study is established by applying convergent and divergent validity methods.
Kaiser-Meyer-Olkin (KMO) and Bartlett’s test for Sphericity are for the confirmation to the items that can be conducted after the former measure.
By using Explanatory Factor Analysis (EFA) and Confirmatory Factor Analysis (CFA), convergent and divergent
validity will be established obviously.
Face Validity
Face validity is a type of validity measurement, which an indicator being reasonable as a measure of a construct when judging others, especially in the scientific area (Neuman, 2011).
Considering the languages that the respondents use, the questionnaire is in English originally and translated into Mandarin.
Content Validity
Content validity is a type of validity measurement which requires a measure that represents all aspects of conceptual definition of a construct (Neuman, 2011). In this study, every items of the questionnaire refers to a definition, and the measures represent their variables as well as dimensions.
Criterion Validity
outside verification (Neuman, 2011). The entire questionnaire is adopted from previous validated instruments, and is examined by pilot test as well.
Construct Validity
Construct validity is a type of validity measurement that applies multiple indicators with two subtypes, including how well the indicators of one construct converge or how well the indicators of different construct diverge (Neuman, 2011). Construct validity will be tested and confirmed by Explanatory Factor Analysis (EFA). Before applying EFA, Kaiser-Meyer-Olkin (KMO) and Bartlett’s test for Sphericity will be accomplished, indicating the necessity and appropriateness of using EFA.
Kaiser-Meyer-Olkin (KMO) Measure of Sampling Adequacy value specifies the appropriateness of applying factor analysis on data, that factor analysis will be considered as useful for those variables and items if KMO value is higher than .80 with an excellent result, since factors extracted from data account for the variance is better when being closer to 1.00 (Friel, 2010). A corrective action might be required if KMO is lower than .50, either to add other factors or items to the variables or to get rid of those that are not essential. In addition, as a remedy of low KMO value, reverse-coding of the negatively-worded items may also be a solution. Bartlett’s test for Sphericity clarifies the factuality of correlations between variables;
by observing Bartlett's test statistic, chi-square value, and p-value (p-value must be larger than .05). Factor analysis can be conducted if the tests are proved with statistics going after the requirements of those values.
Convergent and divergent validity can be explained by using Explanatory Factor Analysis (EFA) and Confirmatory Factor Analysis (CFA). Explanatory Factor Analysis (EFA), applied to verify convergent validity, is responsible to be evidence for correlation between each item’s score and compare the consistency with total score of variables. To be consistent with previous study, the score of one item is expected to be larger than .40 (Kerlinger, 1986).
Furthermore, in order to make clear of divergent validity, Confirmatory Factor Analysis (CFA) is responsible for ensuring the statistics fit the entire threshold well.
Data Analysis Statistical Analysis
The study uses SPSS 22 and SmartPLS software to analyze data, and presents descriptive statistics to express an overview concept of the consequence of the research.
Descriptive Statistics
Without a probabilistic formulation, descriptive statistics are applied to portray the main features of a compilation of data quantitatively (Mann, 1995). It provides a better view to understand the data and where it comes from. To compare the population and sample, categorical variables, for instance, age, gender, educational background and length of employment, etc., frequencies and percentages will be used in this study. Descriptive statistics helps to clarify the considerable sample that represents the case study and identify the influential variable in this investigation. In this study, mean and standard deviation will be used in both pilot and main study.
Correlation Analysis
A correlation analysis is used to explore the direction and strength of linear relationship between variables. The relationships between each variable and independent variable, which might be positive or negative, can be proved by correlation analysis. Nevertheless, it is important to emphasize that a high correlation coefficient may not imply absolutely that multicollinearity existed between the variables. As data being confirmed by CFA, it is also determined as distinct constructs; therefore, if the correlation coefficient value is high but
value of regression can still be accepted.
Regression/Coefficient of Determination (R
2)
Coefficient of determination, or called as regression, will be applied to explain the endogenous latent variables to the total variance. Showing the percentage that how much data gathered can present the real situation of the variables. The higher the percentage is, the closer the data can explain and predict the variables in reality for the study. The structural model will be established by measuring the value of R2 of the latent variables, path coefficients and goodness of fit.
Structural Equation Modeling (SEM)
Also called multivariate or multi-equation method, Structural Equation Modeling (SEM) is a statistical method confirming the relationships between interconnected variables of a research. SEM is usually applied to determine the competency and effectiveness of a model that explains its dimensions appropriately.
Two core SEM techniques are relevant in the study: to perform CFA; and to perform causal modeling, or path analysis, that hypothesizes causal relationships among variables and measures the causal models with a linear equation system. Path coefficients are taken into consider to make clear of the relationship between variables and to determine the direction of the relationships and significance. SEM will be applied to get evidence to support or reject the hypothesis proposed in the beginning.
Partial Least Square (PLS)
PLS path-modeling algorithm is a sort of SEM technique developed by Herman Wold in 1975. Using a combination of principal component analysis, path analysis and a set of regressions, it also estimates path models by latent variable. PLS is an effective tool to predictions and theories, and is supportive for small sample analyses and complicated models.
That is to say, PLS is helpful with exploratory analysis investigating whether relationships might exist among variables (Yu, Kim, & Kim, 2007) and is efficient for estimating causal models in myriad model and data situations.