PLS FINDING - 台積電知識管理與應用

To present the finding results of this chapter was adopted the method by Partial Least Square (PLS). There were three main sections in the pilot study. Comprising pilot result, final result was collected 147 totally questionnaires, and comparison between the respondents from Hsinchu and those from Taichung. By using PLS methodology, this section illustrates more accurate reliability and validity of the items. Besides, R squared shows the explanatory power among all the variables. Finally, the bootstrapping method was used to test the significance of the path coefficients and hypothesis. The mentioned findings and discussions are further explicated.

Pilot Study

This study is conducted the pilot study by collecting 40 questionnaires from participants in headquarter, Hsinchu. After deleted 2 invalid responses, there are 38 convincing questionnaires for successive analysis in this section. To measure the nature of constructs relationships and make the pilot test reliable and valid, this study utilized the PLS method to assess composite reliability, Cronbach alpha, AVE, and factor loading associated with individual items.

The criterions of composite reliability and Cronbach alpha have to be 0.7 to ensure the internal consistency (Nunnally, 1978). In Table5.1, the composite reliability and Cronbach alpha in the most of the constructs are higher than 0.7, except for 0.6044 of Knowledge Management Strategies and 0.6860 of Knowledge Enabler. Generally, all the constructs have high internal consistency, implying each of the devised questions is accurate in directing each construct.

In addition, individual item reliability is also examined the individual item reliability by item loadings. Some researchers accepted the loading of o.5 as the criterion, indicating that there are more shared variance between the constructs (Hulland, 1995; Ringle et. al., 2005).

From table 5.1, individual item loading are much higher than standard excluding 0.2398 of KE_C. Mostly, the item loadings show that the questionnaire should design appropriate items in the questionnaires.

Another measurable indicator is Average Variance Extracted (AVE) 0.5 as its criterion.

It appears constructive when examining the average variance shared in a construct by calculating the correlations with different constructs. All the values reported in Table5.1 are all higher than 0.5 which means all the constructs have the appropriate the relationships between measures and constructs.

Table 5.1.

Pilot Study Variables’ Factor Loadings and Internal Consistency Reliability Analysis via PLS

Constructs

Note: KMS_H=Knowledge Management Strategies of Human; KMS_S=Knowledge Management Strategies of System; KA_E= Experiential Knowledge Assets; KA_C=

Conceptual Knowledge Assets; KA_S= Systemic Knowledge Assets; KA_R= Routine;

KCP_S= Knowledge Creation Process of Socialization; KCP_E= Knowledge Creation Process of Externalization; KCP_C= Knowledge Creation Process of Combination; KCP_I=

Knowledge Creation Process of Internalization; KE_C= Knowledge Enablers of Centralization; KE_T= Knowledge Enablers of Trust; KE_TS= Knowledge Enablers of T shaped skills; KE_IT= Knowledge Enablers of IT support.

In table 5.2, R squared is used to evaluate the extent of explanatory power, resulting in 0.5328 of Knowledge Assets; 0.6690 of Knowledge Creation Process; 0.4303 of Knowledge Enablers; 0.5663 of Performance. Among them, the value of Knowledge Creation Process (KCP) has the highest explanatory power (R²⁼ 0.6690). Besides, the path coefficients of the structural model demonstrate the significance by bootstrapping method. Based on the standardized betas, path coefficients all show the positive directions in table 5.3. However, the one in Knowledge Enablers (KE) have the lowest one (R²⁼ 0.4303), implying the various changes among the distinctive constructs.

Table 5.2.

Pilot PLS Cronbach’s Alpha, Internal Consistency and R2 in This Study

Constructs No. of Items Cronbach's Alpha Internal Consistency R²(%)

KMS 8 0.6044 0.8349 0.0000

KMS KMS_H 0.8454 KMS_S0.8476

KA KA_E 0.6894 KA_C0.6847 KA_S 0.8366 KA_R 0.8929 KCP KCP_S 0.9400 KCP_E 0.8664 KCP_C 0.8776 KCP_I 0.8788

Note: KMS_H=Knowledge Management Strategies of Human; KMS_S=Knowledge Management Strategies of System; KA_E= Experiential Knowledge Assets; KA_C=

Conceptual Knowledge Assets; KA_S= Systemic Knowledge Assets; KA_R= Routine;

KCP_S= Knowledge Creation Process of Socialization; KCP_E= Knowledge Creation Process of Externalization; KCP_C= Knowledge Creation Process of Combination; KCP_I=

By bootstrapping, the significance of the path coefficients of the structural model in Table 5.3 is demonstrated. Based on the standardized betas, all the path coefficients illustrate the positive directions. The figures in the third column of t-value reveal the significance of each path. First, the results display KMS have a strong t-value, counting 8.846 of KA, 1.726 of KCP, and 6.661 of KE. When the probability is smaller than 0.01. The t-value of 11.982 from KCP to P also shows a strongly significant effect. In addition, KA to KCP perform a weak path model relationship (t-value is 2.091). Nevertheless, KE to KCP express no significant effect in this model (t-value is 0.101). Overall, all the hypotheses in the pilot test are for that reason proved supportive except for H5 which is the path from KE to KCP.

Table 5.3.

Pilot PLS Path Analysis Results (Standardized Beta Coefficients and Adjusted T-values)

Path Hypothesis β-path Adj. t-value Sig. Support Direction

KMS  KA H1 0.734 8.846 *** X +

KMS =Knowledge Management Strategies; KA =Knowledge Assets; KCP =Knowledge Creation Process; KE =Knowledge Enablers; P =Performance.

Figure 5.1 SEAP Integrated Model which is created by Shih and Liu demonstrates the result for the pilot study. The results pinpoint that the construct of knowledge management strategies have significant effect on knowledge assets, knowledge creation process, and knowledge enablers. Knowledge assets also have strongly impact on knowledge creation process. However, knowledge enablers have no impact on knowledge creation process. One important benefit of the PLS method is that it makes it possible to separate direct and total effects of all the variables included in this model (Cabrita & Bontis, 2008). That is to say, knowledge creation process can improve the performance through knowledge assets instead of knowledge enablers.

Figure 5.1. SEAP integrated model for pilot study

* p < 0.1. **p <0.05. *** p <0.001.

Table 5.4.shows the null hypothesis results among all the constructs are all rejected except for the hypothesis 5 “Knowledge Enablers have no effect on Knowledge Creation Process” supported. Knowledge enablers have no impact on knowledge creation process that needs to be more discussed in next chapter.

Table 5.4.

Pilot Research Hypothesis Results

Research Hypothesis Results

H1 Knowledge Management Strategies have no effect on the Knowledge Assets

Rejected

H2 Knowledge Management Strategies have no effect on the Knowledge Creation Process

Rejected

H3 Knowledge Management Strategies have no effect on the Knowledge Enablers

Rejected

H4 H5

Knowledge Assets have no effect on Knowledge Creation Process Knowledge Enablers have no effect on Knowledge Creation Process

Rejected Supported H6 Knowledge Creation Process has no effect on Performance Rejected Note:* p < 0.1. **p <0.05. *** p <0.01.

Final Study

In the past months the final result was collected up to 152 online questionnaires provided by the respondents from Hsinchu and Taichung, with a total of 147 effective responses after deleting 5 invalid ones.

Shown in Table 5.5, the composite reliability and Cronbach alpha among all the constructs are higher than the criterion of 0.7 except for 0.6383 of KMS and 0.6875 of KE in the category of Cronbach alpha. Such results imply the low reliabilities of KMS_H and KMS_S in the construct of KMS. Even so, both of their item loadings such as 0.8531 of KMS_H and 0.8607 of KMS_S in this category are above 0.5. The result achieves for the objective of internal consistency through the 4 questionnaires for each. Concerning the factor loading of KE_C, the value of 0.3958 which implies the item of centralization in is the lowest reliability in Knowledge Enablers (KE). Furthermore, all the indicators of the Average Variance Extracted are higher than the criterion of 0.5, highlighting the necessity of examining the average variance shared in a construct and calculating the correlations amid different constructs.

Table 5.5.

Final Study Variables’ Factor Loadings and Internal Consistency Reliability Analysis via PLS

Constructs

Note: KMS_H=Knowledge Management Strategies of Human; KMS_S=Knowledge Management Strategies of System; KA_E= Experiential Knowledge Assets; KA_C= Conceptual Knowledge Assets; KA_S= Systemic Knowledge Assets; KA_R= Routine; KCP_S= Knowledge Creation Process of Socialization; KCP_E= Knowledge Creation Process of Externalization;

KCP_C= Knowledge Creation Process of Combination; KCP_I= Knowledge Creation Process of

Internalization; KE_C= Knowledge Enablers of Centralization; KE_T= Knowledge Enablers of Trust; KE_TS= Knowledge Enablers of T shaped skills; KE_IT= Knowledge Enablers of IT support.

From Table 5.6 below, the result of R-Squared illustrates the good explanatory power in this model due to the fact that the values in most of the constructs are above the criterion of 0.5: 0.5619 of KA, 0.5691 of KCP, and 0.5284 of P. However, the reality that 0.3220 of KE is lower than the average is the same as the result from the pilot test.

Table 5.6.

Final Result of PLS Cronbach’s Alpha, Internal Consistency and R2 in This Study (n=147)

Constructs No. of Items Cronbach's Alpha Internal Consistency R²(%)

KMS 8 0.6383 0.8468 0.0000

KMS KMS_H 0.8531 KMS_S0.8607

KA KA_E 0.7900 KA_C 0.7478 KA_S 0.8227 KA_R 0.8299

KCP KCP_S 0.8597 KCP_E0.8111 KCP_C 0.8381 KCP_I 0.8756 KE

Note: KMS =Knowledge Management Strategies; KA =Knowledge Assets; KCP =Knowledge Creation Process; KE =Knowledge Enablers; P =Performance.

The Standardized Beta Coefficients in Table 5.7 still show the positive directions as same as pilot test, but the results of significance are relatively different from the one in the pilot test. Take KMS as an example, both the adjusted t-values of KA and KE or 7.695 and 3.982 respectively elucidate the strong significance when the probability is lesser than 0.01.

Conversely, there is no significance as the adjusted t-value from KMS to KCP was only 0.176, neither was 1.283 from KE to KCP. When the probability is smaller than 0.10, the adjusted t-value of 2.436 from KA to KCP proves the same result as the one in the pilot study. Therefore, this study demonstrates most of the hypotheses to be helpful exclusive of the one from H2 (KMS to KCP) and H5 (KE to KCP).

Table 5.7.

Final result of PLS Path Analysis (Standardized Beta Coefficients and Adjusted T-values)

Path Hypothesis β-path Adj. t-value Sig. Support Direction

KMS  KA H1 0.750 7.695 *** X +

Figure 5.2 SEAP Integrated Model which is created by Shih and Liu demonstrates the result for this study. The results pinpoint that the construct of knowledge management strategies have significant effect on knowledge assets, knowledge creation process, and knowledge enablers. Knowledge assets also have positive impact on knowledge creation process. However, knowledge enablers have no impact on knowledge creation process as same as pilot study. One important benefit of the PLS method is that it makes it possible to separate direct and total effects of all the variables included in this model (Cabrita & Bontis, 2008). That is to say, knowledge creation process can improve the performance through knowledge assets or knowledge creation process instead of knowledge enablers.

Figure 5.2. SEAP integrated model

* p < 0.1. **p <0.05. *** p <0.001.

Table 5.8 shows the null hypothesis results among all the constructs are all rejected except for the hypothesis 5 “Knowledge Enablers have no effect on Knowledge Creation Process” supported. Knowledge enablers have no impact on knowledge creation process that needs to be more discussed in next chapter.

Table 5.8.

Final Result Research Hypothesis Results

Research Hypothesis Results

H1 Knowledge Management Strategies have no effect on the Knowledge Assets

Rejected

H2 Knowledge Management Strategies have no effect on the Knowledge Creation Process

Rejected

H3 Knowledge Management Strategies have no effect on the Knowledge Enablers

Rejected

H4 H5

Knowledge Assets have no effect on Knowledge Creation Process Knowledge Enablers have no effect on Knowledge Creation Process

Rejected Supported H6 Knowledge Creation Process has no effect on Performance Rejected

Comparisons of Studies from both Hsinchu and Taichung

According to 147 total questionnaires collected from respondents, this study investigates the further result by comparison with the different Science Parks of Hsinchu and Taichung.

The result shows that 81 respondents collected from Hsinchu (55.2 percent) are higher than 66 respondents collected from Taichung (44.8 percent).

In Table5.9, the responses from the participants in Hsinchu demonstrate the composite reliability and Cronbach alpha of all the constructs are higher than 0.7 with the exception of 0.6438 of KMS under Cronbach alpha. In comparison with the results of Hsinchu, in Table 5.10 the reactions from the participants in Taichung show the values of 0.6381 of KMS and 0.6978 of P in the category of Cronbach alpha are lower than the criterions of 0.7, indicating both didn’t achieve the internal consistency.

To further explore the differences between Hsinchu and Taichung, this study evaluates each of the item loadings so as to evaluate the reliable and valid measures of constructs and figure out the nature of the construct relationships among the constructs. The item loadings from Hsinchu in Table 5.9 are all above the criteria of 0.5 excepting 0.3700 of KE_C, while the figures in the same category from Taichung in Table 5.10 are all above 0.5 excluding 0.4052 of KE_C.

Reliability Cronbach Alpha AVE Knowledge

One thing worthy of mentioning is some of the item loadings from the latter emerged lower than the ones from the former, including but not constrained to 0.6564 vs. 0.8048 in KA_C, 0.6113 vs. 0.8477 in KE_T, 0.6884 vs. 0.8461 in KE_TS, 0.7548 vs. 0.8030 in KE_IT, 0.5880 vs. 0.8739 in KCP_E, and 0.6701 vs. 0.8884 in P_F. In fact, it is common to see several measuring items in this model had their loadings below 0.5, particularly when the newly developed scales are used to observation (Hulland, 1999). The reasons of causing low loadings are due to the poorly worded item or the improper transfer of an item from different contexts.

Note: KMS_H=Knowledge Management Strategies of Human; KMS_S=Knowledge Management Strategies of System; KA_E= Experiential Knowledge Assets; KA_C=

Conceptual Knowledge Assets; KA_S= Systemic Knowledge Assets; KA_R= Routine;

KCP_S= Knowledge Creation Process of Socialization; KCP_E= Knowledge Creation Process of Externalization; KCP_C= Knowledge Creation Process of Combination; KCP_I=

In Table 5.11, R squared from Hsinchu shows the value is followed, 0.5616 of Knowledge Assets; 0.6524 of Knowledge Creation Process; 0.3188 of Knowledge Enablers;

0.6159 of Performance. The result shows that R squared of Knowledge Creation Process (KCP) has the highest explanatory power in this model.

Table 5.11.

Hsinchu Result of PLS Cronbach’s Alpha, Internal Consistency and R²in This Study (n= 81)

Constructs No. of Items Cronbach's Alpha Internal Consistency R²(%)

KMS 8 0.6438 0.8475 0.0000

KMS KMS_H 0.8879 KMS_S0.8264

KA KA_E 0.8003 KA_C 0.8048 KA_S 0.8231 KA_R 0.8170 KCP KCP_S 0.8723 KCP_E 0.8739 KCP_C 0.8614 KCP_I 0.8783

In contrast to the results of Hsinchu, the replies from the respondents in Taichung are significantly different. Table 5.12 demonstrates the lowest explanatory power both 0.4868 of KCP and 0.4350 of KE are lower than 0.5.

Table 5.12.

Taichung Result of PLS Cronbach’s Alpha, Internal Consistency and R²in This Study (n= 66)

Constructs No. of Items Cronbach's Alpha Internal Consistency R²(%)

KMS 8 0.6381 0.8447 0.0000

KMS KMS_H 0.8150 KMS_S0.8938

KA KA_E 0.8166 KA_C 0.6564 KA_S 0.8324 KA_R 0.8666 KCP KCP_S 0.8575 KCP_E 0.5880 KCP_C 0.8007 KCP_I 0.8480

Note: KMS_H=Knowledge Management Strategies of Human; KMS_S=Knowledge Management Strategies of System; KA_E= Experiential Knowledge Assets; KA_C=

Conceptual Knowledge Assets; KA_S= Systemic Knowledge Assets; KA_R= Routine;

KCP_S= Knowledge Creation Process of Socialization; KCP_E= Knowledge Creation Process of Externalization; KCP_C= Knowledge Creation Process of Combination; KCP_I=

Knowledge Creation Process of Internalization; KE_C= Knowledge Enablers of

Centralization; KE_T= Knowledge Enablers of Trust; KE_TS= Knowledge Enablers of T shaped skills; KE_IT= Knowledge Enablers of IT support.

The path coefficients of this model demonstrate the significant outcomes by bootstrapping. Based on the standardized betas, path coefficients all show the positive directions among all the path coefficients in Hsinchu in Table 5.13 and in Taichung in Table 5.14. For the former, the t-value in the third column illustrated the significance on each path in Table 5.13. The results shows that KMS have a strongly significance on KA which value is 8.998 and KE which value is 4.331 (***p< 0.01). KA also shows a positive effect on KCP which value is 3.351 (**p< 0.05). In additions, KCP also shows a strongly effect on P which value is 5.760 (***p< 0.01). Conversely, KMS have no effect on KCP which value is 0.281.

Besides, KE also has no significant effect on KCP (0.695). Therefore, the hypotheses of Hsinchu are proved to be supported on H1 (KMS to KA), H3 (KMS to KE), H4 (KA to KCP), and H6 (KCP to P). The hypotheses of Hsinchu are proved to be rejected on H2 (KMS to KCP) and H5 (KE to KCP). The analysis of the significance of t-value indicates that this company in Hsinchu is in need of structuring its comprehensive IT system, developing sufficient T-shaped skills, and expanding adequate managerial institutions.

Table 5.13.

Hsinchu Result of PLS Path Analysis (Standardized Beta Coefficients and Adjusted T-values)

(n= 81)

Path Hypothesis β-path Adj.

t-value

Sig. Support Direction

KMS  KA H1 0.749 8.998 *** X +

Table 5.14 demonstrates the significance of each path in Taichung. KMS have a strongly significance on KA which value is 11.194 and KE which value is 6.435 (***p< 0.01). KA shows a positive effect on KCP which value is 0.836 (***p< 0.01), KE shows a positive effect on KCP which value is 2.593 (**p< 0.05), and KCP have a significant effect on P which value is 4.770 (***p< 0.01). On the contrary, KMS have no effect on KCP which value is 0.195. Besides, KE also has no significance effect on KCP which value is 2.593.

Therefore, the hypotheses of Taichung are verified to be supported with the exception of the hypotheses from H2 (KMS to KCP). The analytical results of t-value in Taichung indicate the authority should consider establishing closer relationships between KMS and KCP.

Table 5.14.

Taichung Result of PLS Path Analysis (Standardized Beta Coefficients and Adjusted T-values) (n= 66)

Path Hypothesis β-path Adj. t-value Sig. Support Direction

KMS  KA H1 0.779 11.194 *** X +

KMS =Knowledge Management Strategies; KA =Knowledge Assets; KCP =Knowledge Creation Process; KE =Knowledge Enablers; P =Performance.

100

Figure 5.3 SEAP Integrated Model which is created by Shih and Liu demonstrates the result for Hsinchu study. The results pinpoint that the construct of knowledge management strategies have significant effect on knowledge assets and knowledge enablers. Knowledge assets also have positive impact on knowledge creation process. However, knowledge enablers have no impact on knowledge creation process as same as final study. One important benefit of the PLS method is that it makes it possible to separate direct and total effects of all the variables included in this model (Cabrita & Bontis, 2008). That is to say, knowledge creation process can improve the performance through knowledge assets instead of knowledge enablers.

Figure 5.3. SEAP integrated model for Hsinchu study

* p < 0.1. **p <0.05. *** p <0.01.

101

Figure 5.4 SEAP Integrated Model shows the result for Taichung. The results pinpoint that the construct of knowledge management strategies have significant effect on knowledge assets and knowledge enablers. However, knowledge assets have no impact on knowledge creation process. it is worth mentioned knowledge enablers have strongly impact on knowledge creation process. One important benefit of the PLS method is that it makes it possible to separate direct and total effects of all the variables included in this model (Cabrita

& Bontis, 2008). That is to say, knowledge creation process can improve the performance through knowledge enablers instead of knowledge enablers.

Figure 5.4. SEAP integrated model for Taichung study

* p < 0.1. **p <0.05. *** p <0.01.

102

Table 5.15 shows the null hypothesis results among all the constructs are all rejected except for the hypothesis 2 “Knowledge Management Strategies have no effect on Knowledge Creation Process” and hypothesis 5 “Knowledge Enablers have no effect on Knowledge Creation Process” supported. The more discussions will be illustrated in next chapter.

Table 5.15

Hsinchu Research Hypothesis Results

Research Hypothesis Results

H1 Knowledge Management Strategies have no effect on the Knowledge Assets

Rejected

H2 Knowledge Management Strategies have no effect on the Knowledge Creation Process

Supported

H3 Knowledge Management Strategies have no effect on the Knowledge Enablers

Rejected

H4 H5

Knowledge Assets have no effect on Knowledge Creation Process Knowledge Enablers have no effect on Knowledge Creation Process

Rejected Supported H6 Knowledge Creation Process has no effect on Performance Rejected

103

Table 5.16 shows the null hypothesis results among all the constructs are all rejected except for the hypothesis 2 “Knowledge Management Strategies have no effect on

在文檔中台積電知識管理與應用 (頁 89-116)