CHAPTER 4 DATA ANALYSIS AND RESULTS
4.1 Descriptive Statistics.…
4.1.2 Characteristics of Samples
The basic information of respondents including their company’s age, the number of employee, capital, their age, seniority, education, department, and management level were described in detail as follows.
It is found that in Table 4-2, 29.4% of participant’s companies are established over 30 years, and 51.6% are founded during 5-20 years. There are 49% companies having employee between 101 and 1000. Moreover, 32% companies’ capitals are from 200 million to 1 billion dollars. From Table 4-3, most of respondents with respect to their age are between 36 to 45 year olds (43.8%). Of the participants, 96% received junior college and graduate school and their service years are mostly 4-9 years (39.8%) and over 12 years (32.7%). In regard to their working divisions, 37.9% respondents are in R&D. About respondents’ management level, 24.2% are in the high level, and 42.5%
are in the middle level.
Table 4-2: Characteristics of the Sampled Firms
Frequency Percent (%)
1 to 4 years 2 1.3
5 to 10 years 24 15.7
11 to 15 years 23 15.0
16 to 20 years 32 20.9
21 to 25 years 12 7.8
26 to 30 years 15 9.8
Company Age
Over 30 years 45 29.4
under 100 persons 14 9.2
101 to 500 persons 51 33.3
501 to 1000 persons 24 15.7
1001 to 2000 persons 14 9.2
2001 to 3000 persons 16 10.5
3001 to 5000 persons 5 3.3
Employee’s Number
Over 5000 persons 29 19.0
Under 10 million dollars 7 4.6 10 to 50 million dollars 6 3.9 50 to 100 million dollars 6 3.9 200 to 500 million dollars 23 15.0 600 million to 1 billion dollars 26 17.0 1 to 2 billion dollars 14 9.2 2 to 5 billion dollars 27 17.6 5 to 10 billion dollars 15 9.8 Capital
Over 10 billion dollars 29 19.0
Table 4-3: Characteristics of the Respondents
Frequency Percent (%)
Under 25 years old 4 2.6
Junior College/ College 83 54.2
Graduate School 64 41.8
Education
No Answer 1 0.7
Manufacture 13 8.5
Marketing 16 10.5
Research and Development 58 37.9
Information 5 3.3
Personnel 10 6.5
Financial 13 8.5
Administrative 19 12.4
Others 16 10.5
4.2 Measurement Results for Research Variables
This research employed the statistic technique of factor analysis by SPSS 15.0 to explore factors that could exist in composites including organizational culture, knowledge creation modes, and creation performance. Besides, reliability and validation of the whole research model are examined. The Structural Equation Modeling (SEM) is adopted to examine the research model fit and to test the hypotheses that are defined in Chapter 2 by LISREL 8.30. Parameter estimation method that is adopted for this research is maximum likelihood. With regarded to the test of the research model fit, measurement model and structural model are used.
Details are described as follows.
4.2.1 Organizational Culture
For the internal consistency reliability, both the item to total correlation and the Cronbach’s α coefficient are used. The higher the α coefficient is, the stronger the interrelationship of each item of measurement as well as the higher internal consistency is. With regard to minimum acceptable criterion, it should be noted that Cronbach’s α that is close to 0.7 is also regarded as an acceptable coefficient (Cuieford, 1965). Also, the item to total correlation is better if the value of it is greater than 0.5. The result was listed in Table 4-4.
For the factor analysis of organizational culture, we used the analysis results of the corrected item-total correlation and Cronbach’s α if item deleted to progressively improve its reliability. Then, Cronbach’s α is the final determinant of the reliability.
culture, reliability for market culture increased. However, although the first and the ninth item for item to total correlation are 0.39 and 0.393, in order to have as much as possible the information presented by the items of the questionnaire, this research kept these two items. Finally, four characteristics, market culture, adhocracy culture, hierarchy culture, and clan culture, explained the total variance of 61.219%.
Table 4-4: Factor Analysis and Reliability Test for Organizational Culture Variable Item Mean Factor
Loading Item to Total Cronbach’s α
1 5.418 0.522 0.390
Note: the total variance explanation is 61.219%
The research then moved to the stage of confirmatory factor analysis (CFA) to investigate the hypothetical casual relationships between independent (organizational culture and knowledge creation modes) and dependent variables (creation performance). The value of χ2 was used to statistically determine the fit of the entire model. Normally, when χ2 is smaller and does not reach the level of significance, the model has a better fit. However, χ2 is quite sensitive to the sample size. The larger the sample size is, the more likely the χ2 reaches the level of significance, and in
consequence the model shows a bad fit (Bentler and Bonett, 1980). Therefore, in order to reduce the sensitivity that may occur to the value of χ2, Hair et al. (1998) suggested that model fit be determined by taking into account every indicator. Hair et al.(1998) proposed three types of overall model fit measures: 1) absolute fit measures, 2) incremental fit measures, and 3) parsimonious fit measures.
An absolute fit measures is used to directly evaluate how well a priori theoretical model fits the sample data (Hu and Bentler, 1995). Measured indicators are as follows：χ2 /d.f., goodness of fit index (GFI), and root mean square residual (RMR).
Bollen (1989) suggested χ2/d.f. less than 3.00 be a good fit of a model. Gefen et al.
(2000) and Hair et al. (1998) suggested that GFI bigger than 0.90 should be a acceptable model fit. Moreover, Hair et al. (1998) proposed RMR between 0.05 and 0.08 is acceptable model.
An incremental fit measures assesses the proportionate amount of improvement in fit when a target model is compared with a more restricted, nested baseline model (Hu and Bentler, 1995). Measured indicators are as follows: normed fit index (NFI), non-normed fit index (NNFI), comparative fit index (CFI), and incremental fit index (IFI). Gefen et al. (2000) and Hair et al. (1998) suggested that those indicators of NFI, NNFI, CFI, and IFI bigger than 0.90 are a acceptable model of fit.
An parsimonious fit measures is used to diagnose whether model fit has been achieved by over-fitting the data with too many coefficients (Hu and Bentler, 1995).
Measured indicators are as follows: parsimonious normed fit index (PNFI), and parsimonious goodness of fit index (PGFI). Byrne (2001) suggested that ideal value
These measures for the variable of organizational culture are described in Table 4-5.
It is found that except GFI that is close to the ideal value, others are acceptable.
Although, the value of χ2(76) is 150.323 and the p-value is 0.000, which shows unacceptable, the values of χ2(76)/df is 1.978, indicating it is less than the ideal value.
Therefore, it is concluded that the model fit of the variable of organizational culture is accepted.
Table 4-5: Model Fit of Organizational Culture Model Fit Indicators Ideal
Value
Analysis Value
χ2 p>0.05 p=0.000
χ2/df <3.0 1.978
GFI(goodness of fit index) >0.90 0.892 Absolute Fit
Measures
RMR(root mean square residual) <0.08 0.0579 NFI(normed fit index) >0.90 0.905
0
NNFI(non-normed fit index) >0.90 0.930 CFI(comparative fit index) >0.90 0.950 Incremental Fit
Measures
IFI(incremental fit index) >0.90 0.950 PNFI(parsimonious normed fit index) >0.50 0.655 Parsimonious
Fit Measures PGFI(parsimonious goodness of fit index) >0.50 0.565
In addition, the confirmatory factor analysis can be used to evaluate the reliability and validation of the model. Measure criteria are 1) absolute standardized coefficient (ASC), 2) composite reliability (CR), and 3) average variance extracted (AVE).
(1) ASC
The ASC is the factor loading that each item is to the potential factors (e.g. MC).
When the t-value of each loading is greater than or equal to 1.96 (α=0.05), 2.58 (α=0.01), or 3.29 (α=0.001), reaching the significant level, the model shows an acceptable reliability (Anderson and Gerbing, 1988).
(2) CR
Similar to the Cronbach’s α, the structure equation modeling (SEM) has been introduced to determine the model reliability by a composite reliability. The CR is the internal consistence measure between individual potential factors and the measured items. Thus, the more the CR is, the more ability it has in measuring the potential items. Hair et al. (1998) suggested that the value of CR be not less than 0.7.
(3) AVE
The AVE is the explanation ability of averaged variance of the measured items for the potential variable. Fornell and Larcker (1981) suggested that the value be not less than 0.5. The larger the AVE is, the better the reliability and validation of the potential variable will be. Moreover, when the AVE of a potential factor is greater than the square of the correlation coefficient of the paired potential factors, the factors analysis result can be regarded as high discriminant validity (Fornell and Larcker, 1981).
The results of model investigation of organizational culture are listed in Table 4-6. In Table 4-7, the statistical measures for the potential factors in the variable of organizational culture are included. It is found that from Table 4-6 and 4-7, based on the validation and reliability investigation criteria mentioned above, the factor loading of each item and the CR and AVE are acceptable. In addition, the AVE of each potential factor is greater than the square of the correlation coefficient of the paired potential factors. By this, it can be seen that the reliability and validity are quite acceptable for the variable of organizational culture.
Table 4-6: Scale Properties of Organizational Culture Variable Item Standardized
Loading T-value Composite Reliability
Table 4-7: AVE of Measured Factors of Organizational Culture
Mean Std. Deviation MC AC HC CC
MC 5.345 0.855 0.849 a 0.585 0.721 0.655 AC 5.162 0.911 0.585 0.900 0.657 0.651 HC 5.277 0.930 0.721 0.657 0.882 0.795 CC 5.198 0.948 0.655 0.651 0.795 0.909 a. Diagonal elements (in bold and italic) in correlation matrix represented the Average Variance Extracted (=
2 2
(Σλ )
[(Σλ )+Σ(θ)]；λ= Factor Loading；θ= Measured Error), while off-diagonal elements were represented by the square of correlation among variables.
4.2.2 Knowledge Creation Modes
For the factor analysis, there is no evidence so far on that this research can rely to measure knowledge creation modes. This research conducted factor exploration and obtained three, including goal-free, goal-depended, and goal-driven. However, in order to increase the internal reliability, item deletion is needed. From Table 4-8, when item 4 and 10 are deleted, the reliability for goal-depended creation mode increased. Yet, although eleventh item for item to total correlation is 0.357, which is very close to 0.4, in order to keep information obtained from the questionnaire as much as possible, this research did not delete these items. However, the factor analysis result did not follow the presumed factors in Chapter 2.We grouped those items and named it as goal-depended (G-DE). Therefore, the final factors are redefined as goal-free (G-FR), goal-depended (G-DE), and goal-driven (G-DR) creation modes. Table 4-8 list these three factors and the related parameters. It is found that the Cronbach’s α for each factor are between 0.65and 0.79. Moreover, the total variance explanation for these factors is 58.215%. Since the factor of G-DE is somehow between G-FR and G-DR, we made further contacts with the participants.
The responses indicate that for some circumstances a company may take on the G-DR mode to effectively frame their strategy, on the one hand, and on the other hand the G-FR to let employees be able to come up with ideas freely. More precisely, with regards to the management level, managers serve as the main decision makers for company’s strategies. In this case, a goal-driven creation mode is usually adopted.
Under a predefined strategy (or a goal) in a higher management level, managers may allow employees to freely look for creative ways to execute or reach the goal.
Therefore, the G-DE creation mode is defined that managers may adopt the goal-driven creation mode for strategic decisions and the goal-free creation mode for
lower level decisions that would support implementation of the defined strategy without losing its focus.
The factor analysis result revealed that the G-DE creation mode. The G-DE would be another mode in this study. The G-DE helps company manager make critical, strategic decision, and then let employees to freely achieve goals. Moreover, the goals could be in the area of product, manufacturing, and management. Although the goal of product creation is predefined, product design and the process of achievement are still free to be achieved and then more creative products would be delivered expectedly. Otherwise, after strategic decision is made, the creative activities with the characteristics of flexible adjustment are helpful to improve the ability of scheduled delivery as well as the ability of product flexibility and output flexibility. By this, the manufacturing creation performance can be enhanced. Management process and management creation performance can be also improved in the similar way.
Therefore, this study also examined whether or not organizational culture has an effect on goal-depended creation mode, and therefore proposed hypotheses H 2-7 to H 2-9.
H 2-7: Goal-depended creation mode has significant influence on product creation performance.
H 2-8: Goal-depended creation mode has significant influence on manufacturing creation performance.
H 2-9: Goal-depended creation mode has significant influence on management creation performance.
Organizational culture with different characteristics could be helpful to strength the adoption of knowledge creation modes. In consequence, this study explored whether or not four culture characteristics (MC, AC, HC, and CC) increase the adoption of
G-DE. Accordingly, this study proposed four hypotheses H 3-9 to H 3-12, as follows:
H 3-9: Market culture has significant influence on goal-depended creation mode.
H 3-10: Adhocracy culture has significant influence on goal-depended creation mode.
H 3-11: Hierarchy culture has significant influence on goal-depended creation mode.
H 3-12: Clan culture has significant influence on goal-depended creation mode.
Table 4-8: Factor Analysis and Reliability for Knowledge Creation Modes
Variable Item Mean Factor Loading
Item to
Total Eigenvalues Explained
Variance Cronbach’s α
Note: the total variance explanation is 58.215%
For the model fit of knowledge creation modes, it is found from Table 4-9 that χ2(44) is 108.600 and p-value is 0.000, indicating not acceptable. However, the χ2/df is 2.468, indicating acceptable. Also, the GFI is 0.896, NFI is 0.863, NNFI is 0.867, and RMR is 0.0971, indicating close to the ideal value. Other parameters are all accepted.
Accordingly, the model fit is acceptable for the knowledge creation modes.
Regarding the reliability and validation, it is found from Table 4-10 and 4-11 that factor loading is significant and CR and AVE are more than the ideal value. Moreover, all AVE of potential factors are greater than the square of related coefficient of the
modes are acceptable.
Table 4-9: Model Fit for Knowledge Creation Modes Model Fit Indicators Ideal
Value
Analysis Value
χ2 p>0.05 p=0.000
χ2/df <3. 2.46
GFI(goodness of fit index) >0.90 0.896 Absolute Fit
Measures
RMR(root mean square residual) <0.08 0.0971 NFI(normed fit index) >0.90 0.863
00 8
NNFI(non-normed fit index) >0.90 0.867 CFI(comparative fit index) >0.90 0.911 Incremental
Fit Measures
IFI(incremental fit index) >0.90 0.914 PNFI(parsimonious normed fit index) >0.50 0.575 Parsimonious
Fit Measures PGFI(parsimonious goodness of fit index) >0.50 0.505
Table 4-10: Scale Properties for the Knowledge Creation Modes Variable Item Standardized
Loading T-value Composite Reliability
Table 4-11: AVE of Measured Factors of Knowledge Creation Modes Mean Std. Deviation G-DR G-FR G-DE
G-DR 4.132 1.087 0.821 a 0.226 0.287 G-FR 5.187 0.742 0.226 0.710 0.472
G-DE 4.956 0.914 0.287 0.472 0.852 a. Diagonal elements (in bold and italic) in correlation matrix represented the Average
Variance Extracted (=
2 2
(Σλ )
[(Σλ )+Σ(θ)]；λ= Factor Loading；θ= Measured Error), while off-diagonal elements were represented by the square of correlation among variables.
4.2.3 Creation Performance
With regard to the factor analysis for creation performance, the results are obtained in Table 4-12. It is found that Cronbach’s α for each factor is between 0.91 and 0.94.
Factors that can be confirmed are product creation performance (PCP), manufacturing creation performance (MFCP), and management creation performance (MGCP). Moreover, the total variance explanation for these factors is 69.595%. Also, the result obtained for test model fit is shown in Table 4-13. It is found that χ2(195) is 493.607 and p-value is 0.000, indicating not acceptable. However, the χ2/df is 2.531 that is less than the ideal value, indicating acceptable. The value of GFI is 0.807 and NFI is 0.880, indicating close to the ideal value. Other parameters are all acceptable.
Therefore, the model fit for the variable of creation performance is acceptable. For the investigation of reliability and validation, it is found from Table 4-14 and 4-15 that factor loading is significant and CR and AVE are more than the ideal value.
Moreover, all AVE of potential factors are greater than the square of related coefficient of the paired factors. Thus, reliability and validation of the variable of knowledge creation modes are acceptable.
Table 4-12: Factor Analysis and Reliability Test for Creation Performance Variable Item Mean Factor
Loading
Item to
Total Eigenvalues Explained
Variance Cronbach’s α
Note: the total variance explanation is 69.595%
Table 4-13: Model Fit for the Creation Performance Model Fit Indicators Ideal
Value Value
χ2 p>0.05 p=0.000
χ2/df <3. 2.
GFI (goodness of fit index) >0.90 0.807 Absolute Fit
Measures
RMR (root mean square residual) <0.08 0.0652 NFI (normed fit index) >0.90 0.880
00 531
NNFI (non-normed fit index) >0.90 0.900 CFI (comparative fit index) >0.90 0.923 Incremental
Fit Measures
IFI (incremental fit index) >0.90 0.924 PNFI (parsimonious normed fit index) >0.50 0.679 Parsimonious
Fit Measures PGFI (parsimonious goodness of fit index) >0.50 0.571
Table 4-14: Scale Properties for the Measurement Model of Creation Performance Variable Item Standardized Loading T-value Composite
Reliability
Table 4-15: AVE of Measured Factors of Creation Performance
Mean Std. Deviation PCP MFCP MGCP
PCP 5.279 0.926 0.783 a 0.711 0.478
MFCP 5.451 0.855 0.711 0.775 0.580
MGCP 5.211 0.929 0.478 0.580 0.801
a. Diagonal elements (in bold and italic) in correlation matrix represented the Average Variance Extracted (=
2 2
(Σλ )
[(Σλ )+Σ(θ)]；λ= Factor Loading；θ= Measured Error), while off-diagonal elements were represented by the square of correlation among variables.
4.2.4 Research Model
For the entire research model fit, the confirmatory factor analysis is used to test the causal relationships between independent variables and dependent variable. The results obtained in Table 4-16 indicate that χ2(32) is 130.763 and p-value is 0.000, which is not acceptable. However, the χ2/df is 2.255, showing acceptable because it is less than the ideal value. By this, it is found that the whole research model fit is acceptable. In addition, the results shown in Table 4-17 and 4-18 indicate that based on the test criteria of reliability and validation, the factor loading of each item is significant. The value of both CR and AVE are greater than the ideal value. Moreover, all AVE of potential factors are greater than the sum of related coefficient of the paired factors. In short, the reliability and validation of the whole research model is acceptable.
Table 4-16: The Fit of the Research Model
GFI(goodness of fit index) >0.90 0.922 Absolute Fit
Measures
RMR(root mean square residual) <0.08 0.0558 NFI(normed fit index) >0.90 0.927
0 5
NNFI(non-normed fit index) >0.90 0.936 CFI(comparative fit index) >0.90 0.957 Incremental
Fit Measures
IFI(incremental fit index) >0.90 0.958 PNFI(parsimonious normed fit index) >0.50 0.618 Parsimonious
Fit Measures PGFI(parsimonious goodness of fit index) >0.50 0.503
Table 4-17: Scale Properties of the Research Model Construct/Variable Standardized
Loading T-value Composite Reliability
Table 4-18: AVE for Research Variables
Mean Std. Deviation OC KCM CP
OC 5.245 0.794 0.903 a 0.561 0.722
KCM 10.817 1.728 0.561 0.795 0.419
CP 5.314 0.769 0.722 0.419 0.859
a. Diagonal elements (in bold and italic) in correlation matrix represented the Average Variance Extracted (=
2 2
(Σλ )
[(Σλ )+Σ(θ)]；λ= Factor Loading；θ= Measured Error), while off-diagonal elements were represented by the square of correlation among variables.
4.3 Structural Equation Modeling
In this research, in order to ensure the research model is acceptable, the structural equation model is developed both to examine the model fit and to test the defined hypotheses. For the model fit, it is found from Table 4-19 that χ2(30) is 67.650, indicating not acceptable. However, the value of χ2/df is 2.255, indicating acceptable because it is less than the ideal value. Other parameters are all acceptable. By this, it is found that the whole research model fit is acceptable.
Table 4-19: Model Fit for the Structure Equation Model Model Fit Indicators Ideal
Value
Analysis Value
χ2 p>0.05 p=0.000
χ2/df <3.0 2.255
GFI(goodness of fit index) >0.90 0.922 Absolute Fit
Measures
RMR(root mean square residual) <0.08 0.0445 NFI(normed fit index) >0.90 0.927
0
NNFI(non-normed fit index) >0.90 0.936 CFI(comparative fit index) >0.90 0.957 Incremental
Fit Measures
IFI(incremental fit index) >0.90 0.958 PNFI(parsimonious normed fit index) >0.50 0.618 Parsimonious
Fit Measures PGFI(parsimonious goodness of fit index) >0.50 0.503
For the hypothesis test, structural equation is built to conduct the path analysis to examine whether or not the directions and causal relationships are significantly accepted. The test criteria used include 1) standardized path coefficient which shows causal relationship (direction) between independent variables and dependent variable, and 2) determination coefficient (R2). For the former, the standardized path coefficient measures the path coefficient and regression weight between latent variables. That is, it measures the strong extent of casual relationship between variables. The goal is to check whether or not the path direction of the model is the same as that defined in the research model. If the estimated parameter goes to a positive value, the research model meets the direction. If it shows negative, the predefined path does not meet what the collected data says. Significance is determined by the t value. For example, it is significant if t is greater than or equal to 1.96 under the significance level of 0.05 (α=0.05). For the later, the R2 represents the variance of each internal variable. It is the percentage of explanation of causal relationship by its independent variable. The larger the R2 is, the better the equation model will be. The path diagram and its parameters are shown in Figure 4-1.
Knowledge Creation Modes R2=0.775
Organizational Culture Creation Performance R2=0.801 0.880***
(4.05)
0.308 (1.33)
Goal-Depended Goal-Driven Goal-Free
0.765*** 2. － represent significant
3. --- represent insignificant 4. ( ) represent t value
Figure 4-1: Path Diagram of SEM Model
It is found from Figure 4-1 that all path directions match those defined in the research model. In addition, parameters of the path from knowledge creation modes to creation performance indicate insignificant (standardized path coefficient = 0.308, t = 1.33)
It is found from Figure 4-1 that all path directions match those defined in the research model. In addition, parameters of the path from knowledge creation modes to creation performance indicate insignificant (standardized path coefficient = 0.308, t = 1.33)