• No se han encontrado resultados

4. PARÁMETROS DE IMPLEMENTACIÓN Y UBICACIÓN

4.3. RESULTADOS

This section describes the statistical techniques chosen for this study to answer the four research questions addressed in this research. This will be described below.

4.10.1 Exploratory Factor Analysis

Exploratory factor analysis (EFA) was undertaken to explore and classify the best items that can represent the constructs under study. This Pallant (2007, p.179) suggests is “looking for a way to summarise the large data becoming

a smaller set of factors or components”. It was therefore, necessary to

identify the structure of variables that contributed to the construct in order to avoid problems in interpretation of the extent to which each variable may affect outcomes (Field, 2005). Since the scales used to assess perceptions of organisational culture and perceived level of knowledge sharing capability combined measures from a number of different studies, it was necessary to confirm their dimensionality empirically. Thus, in this research, exploratory factor analysis was used to confirm the dimensions of the concepts that have been operationally defined as well as to indicate which of the items were most appropriate for each dimension (Hair et al., 2010; Spector, 1992). Factor analysis was undertaken using PASW version 18. The procedures undertaken were explained below.

121 | P a g e 4.10.1.1 Identifying Univariate Outliers

Each item for each variable in the conceptual framework was checked using z-scores in order to detect any case that had values more than ±3 standard deviations from the mean of the variables (Tabachnick& Fidell, 2007). If any outlier was found, the case with the outlier was removed from the analysis. In this research, no outliers were found; therefore none of the cases was removed from the data set.

4.10.1.2 Accessing the Characteristics of Matrices

Next, the correlation matrix was examined in order to check the factorability among the items, and the strength of the inter-correlations among the items for evidence of coefficients greater than 0.3 (Pallant, 2007; Tabachnick & Fidell, 2007). Factor analysis may not be appropriate if few correlations above 0.3 are found (Pallant, 2007). In examining the characteristics of matrices, Pett, Lackey, & Sullivan’s, (2003) suggestion was followed. The following steps were undertaken:

x Examining the correlation matrix. If there are many coefficients greater than 0.3, the determinant of a matrix would then be evaluated. According to Pett et al. (2003) the determinant of a matrix is a unique numeric value that is associated only with square matrices and is critical for determining whether or not a given square matrix will have an inverse (Hays, 1994 cited in Pett et al., 2003). This relationship is essential to the undertaking of mathematical operations of matrices (Pett et al., 2003). When insufficiently strong correlations are found among the items, then the poorly correlated item is dropped from the analysis and the matrix re-run.

x Evaluating the determinant of a matrix. If the determinant |R| value is equal to 0.000 the correlation matrix is a singular matrix, not positive definite, suggesting that some items are too highly correlated and dropping the highly correlated items is recommended. If, however, the

122 | P a g e determinant |R| value is equal to 1.0, the correlation matrix is an identity matrix, suggesting that the factor analyses are inadvisable. The determinant |R| value should be bigger than 0.0 and smaller than 1.0 (0.0 < |R| <1.0) if the data are suitable for factor analyses be undertaken (Pett et al., 2003).

x Examining Bartlett’s test of Sphericity. Bartlett’s test of Sphericity

shows whether the correlation matrix is an identity matrix, which would indicate that the factor model is inappropriate (Pett et al., 2003). The

Bartlett’s test of Sphericity should be statistically significant at p <

0.05, suggesting a sufficient minimum sample size and the correlation matrix is not an identity matrix. If the Bartlett’s test of Sphericity is not significant at p < 0.05, it indicates the sample size is insufficient relative to the number of items. It is advisable then to increase the sample or reduce the number of items and re-run the matrix, or else factor analyses are inadvisable.

x Examining the Kaiser-Meyer-Olkin (KMO) and individual Measures of Sampling Adequacy (MSA) value. The KMO statistic indicates the sufficient sample size relative to the number of items in the scale and MSA suggests whether or not the correlations among the individual items are strong enough to indicate the correlation matrix is factorable. The minimum recommended value for both KMO and individual MSA are 0.60 (KMO > 0.6) and 0.70 (MSA > 0.70) respectively (Hair et al., 2010; Pallant, 2007; Pett et al., 2003). According to Hair et al. (2010), the MSA values for individual items should be examined to identify potentially problematical items and to eliminate any that did not meet the minimum recommended value. If there are any, they will be eliminated from factor analysis one at a time, with the smallest one being omitted each time and a new matrix solution that excludes the eliminated items should then be undertaken and the results revaluated (Hair et al., 2010; Pett et al., 2003).

123 | P a g e

4.10.2 Analysis of Variance (ANOVA)

Analyses of variance (ANOVA) were undertaken to examine the impact of subgroups of employees (management versus non-management) on the level of knowledge sharing capability, organisational culture and knowledge sharing success. The aim was to decide whether or not further analysis should distinguish respondents by subgroups of employees.

4.10.3 Correlations

Correlation analyses using the Pearson were undertaken between all constructs in the model to identify significant correlations existing between the constructs. When examining the strength of the relationship between all constructs, the researcher used the guidelines provided by Cohen (1988):

r = 0.10 to 0.29 small r = 0.30 to 0.49 medium r = 0.50 to 1.00 large

Significant correlations between the dependent and independent variables lent support to the use of regression analysis as the next step. Details of the findings on correlation analyses were discussed in the findings chapter (Chapter 5).

4.10.4 Regression Analysis

Multiple regression analyses were also conducted to identify significant relationships existing between the variables (perceptions of knowledge sharing capability; perceptions of organisational culture; and perceptions of knowledge sharing success) presented in the conceptual framework.

124 | P a g e

4.11 CHAPTER CONCLUSION

This chapter has provided an overview of the approached taken in undertaking the research. A multiple items questionnaire was used to measure the variables mentioned in the conceptual framework. Sampling methods and participant demographics were also provided along with an explanation of the creation of constructs for investigation, including procedures undertaken for data collection, data analysis strategy and both reliability and validity testing. A sample of 500 knowledge workers of Malaysian-owned IT organisations with MSC status that employed more than 100 personnel, located in two states of Malaysia were chosen. The statistical techniques used to address the research questions include factor analysis, ANOVA, Pearson correlations, and multiple regression analyses. The next Chapter will describe the findings of this research.

125 | P a g e

Documento similar