Empirical Assessment of the Causal Model
THE RELATIONSHIP AMONG THE FIVE SISP FACTORS was examined using structural
equation modeling within the LISREL VII framework [32] .* LISREL is a second-generation,
multivariate technique that can be used to estimate the parameters of causal
models. Second-generation techniques recognize two components of a causal model:
the measurement model and the structural model. The measurement model consists
of the relationships between the constructs and the items used to measure them. It
recognizes that observed measurements are only indirect estimates of the actual constructs,
which are latent or unobservable. The structural model consists of the unobservable
constructs and the theoretical relationships (i.e., the paths) between them.
Thus, LISREL involves constructing a model, estimating its parameters from the
data, and testing the fit of the model to the data by comparing the observed correlations
with the correlations among the variables predicted by the model [17]. LISREL can
handle errors in measurement, correlated errors and residuals, and reciprocal correlations.
It uses maximum likelihood estimation, a full infonnation approach in which it
estimates all parameters simultaneously.
The SISP factors in this study represented latent variables measured by the items shown
in Table 3. Figure 1 shows the hypothesized causal relationship among the factors.
Organization was hypothesized as an exogenous variable (variables that cause other
variables and are assumed to be determined by causes outside the model). Hardware,
Database, Cost, and Implementation were hypothesized as endogenous variables.
The structural equation model used for the analysis is written as: