Construct validity is the ability of a measure to confirm a network of related hypothe- ses generated from a theory based on constructs. Internal construct validity was as- sessed using factor analysis. Because factor analysis provides evidence of the dimen- sion of a measure, factor analysis with a Varimax rotation was used to determine the number of factors contained in each dimension. An eigenvalue greater than 1 is consid- ered to indicate the presence of an inter-pretable factor (Kaiser, 1958); therefore, factors with eigenvalues greater than 1 were taken into account for further analysis. Such rule is the default one used by SPSS unless another one is specified (Stevens, 2002 recited in Hoang Trong and Chu Nguyen Mong Ngoc (2008).
Construct validity was further evaluated through measuring convergent validity, which refers to the extent to which: (i) different scales of constructs indicate the same dimen- sion; and (ii) multiple measures of the same construct are matched (Kerlinger, 1986 re- cited in Hoang Trong and Chu Nguyen Mong Ngoc (2008)). Convergent validity was checked to ensure that each group of constructs indicates the same dimension, and
to verify the degree of compatibility among multiple measures within the same con- struct (Kerlinger, 1986 recited in Hoang Trong and Chu Nguyen Mong Ngoc (2008).
Convergent validity exists “when measures of the same concept have similar patterns of correlation with other variables” (Weisberg et al., 1996 recited in Hoang Trong and Chu Nguyen Mong Ngoc (2008). Construct validity was evaluated by following guide- lines for measuring convergence proposed by Bagozzi, 1981 (recited in Hoang Trong and Chu Nguyen Mong Ngoc (2008)). Bagozzi, 1981 states that correlations for items within a dimension should be high. Convergent validity was assessed by measuring the correlation among the corresponding constructs under each of the four dimensions:
(i) strategy; (ii) processes; (iii) technology; and (iv) people. High correlations among constructs under each dimension are considered to indicate convergent validity. Ex- istence of convergent validity is determined if all correlations between constructs are higher than 0.5 (Liu, 2001).
The number of factors is determined based on the Eigenvalue index, this index repre- sents the variance explained by each factor. According to the Kaiser criteria, the index Eigenvalue factor which is smaller than 1 is excluded from the model (Garson,2003).
Variance explained criteria: the total of variance explained should be greater than 50%
(Hair et al., 1998 recited in Hoang Trong and Chu Nguyen Mong Ngoc (2008). In order for the measurement scale to receive a convergence value, the correlation coefficient between the variables and the factor loading must be greater than or equal to 0.5 in a factor. Principal component analysis with varimax rotation must be done to ensure that the number of factors is minimum.
EFA is considered to match to the set of data if it satisfies the following criteria: First, the matching between EFA and sample data is verified by Kaise-Meyer-Olkin (KMO) statistical values. In which, if the value of KMO is greater than 0.5, the EFA is ap- propriate (Garson, 2003recited in Hoang Trong and Chu Nguyen Mong Ngoc 2008), otherwise if the value of KMO is less than 0.5, EFA is not suitable for the collected
data. Second, number of factors: The number of factors is determined based on the eigenvalue index representing the variation portion explained by each factor. Accord- ing to Kaiser’ criteria, the factor whose eigenvalues are less than 1 is removed from the research model (Garson, 2003). Third, variance explained criteria: Sum of vari- ance explained criteria must be greater than 50% (Hair et al., 1998 recited in Hoang Trong and Chu Nguyen Mong Ngoc (2008). Fourth, convergence value: To ensure the convergence of scales, the single correlation coefficients between variables and factor loading must be greater than or equal to 0.5 in one factor (Gerbing and Anderson, 1988 recited in Hoang Trong and Chu Nguyen Mong Ngoc (2008). Finally, principal com- ponent analysis with Varimax rotation is used to ensure that the number of factors is minimized Hoang Trong and Chu Nguyen Mong Ngoc (2008).
The results of KMO test for e-readiness and perceived factors are presented in table 3.4. The Rotated Component Matrix for e-readiness measure and perceived measure are shown in AppendixC.
TABLE3.4: KMO and Bartlett’s Test
KMO and Bartlett’s Test for E-readiness (27 items) KMO Measure of Sampling Adequacy .932 Bartlett’s Test Approx. Chi-Square 5325.88
Bartlett’s Test df. 351
Bartlett’s Test Sig. .000
KMO and Bartlett’s Test for Perceived Factors (20 items) KMO Measure of Sampling Adequacy .905 Bartlett’s Test Approx. Chi-Square 2160.98
Bartlett’s Test df. 190
Bartlett’s Test Sig. .000
.
As can be seen from table3.4, KMO coefficient for 27 items in e-readiness measurement equals to 0.932, ensures the requirements that 0.5 < KMO <1; with significant level Sig. of 0.000 meets the Sig. less than 0.005 requirement. Furthermore, varimax rotated component matrix for e-readiness measure, as can be seen from appendix C, all 27
items in the questionnaire have factor loading more than 0.30 and they are remained in the model.
Similarly, the KMO and Bartlett’s test result for 20 questions in perceived factors, as can be seen from table3.4KMO for POER and PEER is .905 with a significant of .0000 meeting the Sig. less than 0.005 requirement. In addition to that, AppendixCshows that items 1 through 14 have a loading factor bigger than 0.5 and should be grouped into factor 1 and named POER while items 15 through 20 should be grouped into the other factor called PEER.