**6. Results**

**Sections No. of items**

**Figure 6.** KS tools not used at all among knowledge workers in MSC organizations.

vey took place. A total of 30 respondents participated in these tests.

Most of these organizations are in their initial stages of tool implementation or tools have been implemented but with very minimum tools usage among knowledge workers in these organizations. This rationalized the importance to carry out a research on their intention to

165

235

0 100 200 300 400 500

234

307

116

Comparative Analysis on Frequency of different KS tools usage - Not at All

> 70 65

45

403

All the constructs in the research model were measured with items adapted from prior research. All the items in the questionnaire used a five-point Likert-type scale ranging from

Respondents were asked to indicate to what extend one feels when one evaluates the KS tools when interacting and using the tools. The respondents were asked to record their feelings that were induced by the tools when they interacted with them. The positive and negative affect (PA and NA) are hypothesized to have an impact on PEOU, PU, and BI in the research model. Items for positive and negative affect (PANAS) were adapted from Perlusz [39], Tellegen [28], Watson [43], and Watson et al. [44]. Pre-test and pilot test were carried out before actual sur-

The proposed model and hypothesis testing was carried out using SmartPLS 3.0 software. The measurement and structural model analysis follows methodology described in Hair

General information 9 KS tools behavioral intention 30 Role of affect 24

"Strongly Agree" to "Strongly Disagree" (**Table 1**).

use KS tools in their day-to-day jobs.

Video Sharing

Discussion Forum

156 Knowledge Management Strategies and Applications

WebMeeting

Social Media

Blogs

**5.2. Instrument and measures**

**Table 1.** Instrument.

**5.3. Data analysis**

### **6.1. Measurement model analysis**

This section discusses the measurement model, which consists of several analyses. **Table 2** illustrated the composite reliability on the results. Based on the analysis, it was shown that PU, PEOU, ATT, and BI all achieved the value of composite reliability higher than 0.90, which satisfy the threshold of composite reliability.

Convergent validity looks at the extent to which a measure correlates positively with alternative measures of the same construct. In **Table 2**, it was found that all the constructs' AVE are significant with at least 0.50 and above.

Discriminant validity is the extent to which a construct is truly distinct from other constructs by empirical standards. This means a construct has captured the phenomena not represented by other constructs in a model. Cross loading and Fornell-Larcker criterion discriminant validities are used in this analysis. The analysis indicates that AVE of all constructs has high correlation. Formative measurement analysis conducts a separate set of validity and consistency test. The formative constructs in the proposed model consist of positive and negative affect for PU (NA\_PU and PA\_PU), positive and negative affect for PEOU (NA\_PEOU and PA\_PEOU), and positive and negative affect for BI (NA\_BI and PA\_BI). Convergence validity is the extent to which a measure correlates positively with other measures or indicators of the same construct. A redundancy analysis is used to perform the convergence validity test by


**Table 2.** Composite reliability for reflective constructs.

evaluating formative measurement models; one must test whether the formatively measured construct is highly correlated with a reflective measure of the same construct. A global indicator is designed for this test. To conduct the convergent validity, a separate model is created with the global indicator for each formative construct. From the outcomes of the redundancy analysis, negative affect on perceive usefulness, negative affect on perceive ease of use, negative affect on behavioral intention, positive affect on perceive ease of use and positive affect on behavioral intention are 0.8 or above. Their formative indicators are significant enough to capture content that these constructs want to capture.

For collinearity assessment, when collinearity has high correlations between two formative indicators in a formative construct, it is a problematic indicator and it is unwanted for a formative construct. VIF is used to assess collinearity. Once the collinearity of formative indicators has been treated, outer weights in formative measurement models can then be interpreted. All formative indicators satisfy the requirement of VIF values uniformly with values below the threshold value of 5. There are five items used to test PA → PU, PA→PEOU, and PA → BI. Similarly, five items were designed for NA → PU, NA → PEOU, and NA → BI. The items that are labeled as AA1A..E, BB1A..E, and CC1A..E are positive affect items whereas AA1F..I, BB1F..I and CC1F..I are negative affect items. There is no collinearity problem found in the model except for the items AA1G and AA1H from negative affect on perceived usefulness, BB1G and BB1H from negative affect on perceived ease of use and CC1F and CC1J from negative affect on behavioral intention. Based on items AA1G, AA1H, BB1G, BB1H, BB1J, CC1F, and CC1J being important questions to measure negative affect on perceived usefulness, perceived ease of use, and behavioral intention in the instrument, therefore, these items will be retained. A formative indicator of its relevance is analyzed based on the values of its outer weight as it is compared with others to determine its relative contribution to the construct.

To determine whether an indicator is significant or not, each indicator's *t*-value must fulfill the critical value of l.65 for two-tailed tests at a significant level = 10%. The indicator significance level analysis for positive affect and negative affect on perceived usefulness is significant. Positive and negative affect on perceived ease of use has ten formative items that measure the construct. Five items were chosen to measure positive and negative affect, respectively. One negative affect item is not significant. Its *t*-value is less than 1.65, and outer weight and outer loading do not fulfill the criteria. However, BB1J was not considered to be deleted from this construct because this item has been validated and tested in the previous instrument. Positive and negative affect on behavioral intention has ten formative items that measure the construct. Five items were chosen to measure positive and negative affect, respectively. Negative affect has one item that is not significant. The outer loading value of 0.482 for CC1J is rounded up to be 0.5. Hence, all items are significant. In short, based on the theoretical model and measurement scale used for the proposed research model, the existing items for each construct will be kept.

### **6.2. Structural model analysis**

Collinearity assessment on a structural model involves examination of each set of predictor constructs for each part of the structural model. Collinearity is assessed based on those constructs that have tolerance levels below 2.0 or VIF above 0.50. If such collinearity exists, one should consider eliminating the constructs, merging predictors into a single construct, or creating higher-order constructs to treat collinearity problem. **Table 3** shows that there is no collinearity problem encountered in the research model.

Structural model is used to calculate the estimates of the structural model relationships (path coefficient) that are represented as the hypothesized relationships among the constructs. For this research, we choose to take a significant level of 10% with a critical value of 1.65. Besides examining *t*-values, *p*-values are considered in this analysis. To obtain the *t*-values, a bootstrapping procedure with 5000 resamples was applied. Based on the analysis results, the hypothesis testing results are summarized as follows in **Table 4**.

Another important measure is the total effect of each path. Direct effect for each path may not be very significant in some cases; hence, Total Effect is to assess the significant of paths in the model. The coefficient of determination is a measure of the model's predictive accuracy using *R*<sup>2</sup> where it represents the exogenous latent variables' combined effects on the endogenous latent variables. *R*<sup>2</sup> also represents the amount of variance in the endogenous constructs explained by all the exogenous constructs linked to it. In scholarly research that focuses on marketing issues *R*<sup>2</sup> values of 0.75, 0.50, and 0.25 for endogenous latent variables can be described as substantial, moderate, and weak, respectively.

Attitude toward using KS tools can predict with an accuracy that is close to value 1. Followed by behavioral intention to use KS tools with a *R*<sup>2</sup> value of 0.625 and Task Category-KS tools Fit of a *R*<sup>2</sup> value of 0.593. As for perceived usefulness, it has a *R*<sup>2</sup> value of 0.45 and perceived ease of use has the smallest *R*<sup>2</sup> value of 0.360. By examining *t*-values based on the critical values 1.65 for two-tailed tests at a significant level = 10%, all the *t*-values in the table are significant. Hence, all the predictive accuracy values are significant (**Table 5**). Hence, ATT\_G and BI\_I are substantial and PU\_I and PEOU\_H are moderate endogenous latent variables in the proposed model.


**Table 3.** Summary of VIF for collinearity analysis.

evaluating formative measurement models; one must test whether the formatively measured construct is highly correlated with a reflective measure of the same construct. A global indicator is designed for this test. To conduct the convergent validity, a separate model is created with the global indicator for each formative construct. From the outcomes of the redundancy analysis, negative affect on perceive usefulness, negative affect on perceive ease of use, negative affect on behavioral intention, positive affect on perceive ease of use and positive affect on behavioral intention are 0.8 or above. Their formative indicators are significant enough to

For collinearity assessment, when collinearity has high correlations between two formative indicators in a formative construct, it is a problematic indicator and it is unwanted for a formative construct. VIF is used to assess collinearity. Once the collinearity of formative indicators has been treated, outer weights in formative measurement models can then be interpreted. All formative indicators satisfy the requirement of VIF values uniformly with values below the threshold value of 5. There are five items used to test PA → PU, PA→PEOU, and PA → BI. Similarly, five items were designed for NA → PU, NA → PEOU, and NA → BI. The items that are labeled as AA1A..E, BB1A..E, and CC1A..E are positive affect items whereas AA1F..I, BB1F..I and CC1F..I are negative affect items. There is no collinearity problem found in the model except for the items AA1G and AA1H from negative affect on perceived usefulness, BB1G and BB1H from negative affect on perceived ease of use and CC1F and CC1J from negative affect on behavioral intention. Based on items AA1G, AA1H, BB1G, BB1H, BB1J, CC1F, and CC1J being important questions to measure negative affect on perceived usefulness, perceived ease of use, and behavioral intention in the instrument, therefore, these items will be retained. A formative indicator of its relevance is analyzed based on the values of its outer weight as it is compared with others to determine its relative contribution to the construct.

To determine whether an indicator is significant or not, each indicator's *t*-value must fulfill the critical value of l.65 for two-tailed tests at a significant level = 10%. The indicator significance level analysis for positive affect and negative affect on perceived usefulness is significant. Positive and negative affect on perceived ease of use has ten formative items that measure the construct. Five items were chosen to measure positive and negative affect, respectively. One negative affect item is not significant. Its *t*-value is less than 1.65, and outer weight and outer loading do not fulfill the criteria. However, BB1J was not considered to be deleted from this construct because this item has been validated and tested in the previous instrument. Positive and negative affect on behavioral intention has ten formative items that measure the construct. Five items were chosen to measure positive and negative affect, respectively. Negative affect has one item that is not significant. The outer loading value of 0.482 for CC1J is rounded up to be 0.5. Hence, all items are significant. In short, based on the theoretical model and measurement scale used for the proposed research model, the existing items for

Collinearity assessment on a structural model involves examination of each set of predictor constructs for each part of the structural model. Collinearity is assessed based on those

capture content that these constructs want to capture.

158 Knowledge Management Strategies and Applications

each construct will be kept.

**6.2. Structural model analysis**


**Table 4.** Summary of the hypothesis testing results.


**Table 5.** *R*<sup>2</sup> .
