Data Analysis
Overview of Statistical Analysis
The SPSS and LISREL 8.30 (Joreskog & Sorbom, 1996) were used to conduct all
statistical analyses. The criterion for significance tests for all a priori hypotheses was set
at c =.05. To test the first hypothesis, the relationship between reading fluency and four
latent variables, structural equation modeling with latent variables was used. To test
Hypotheses 2 through 5, a regression analysis with SEM was used to examine the
relationship among four independent variables and one dependent latent variable. Finally,
to test Hypotheses 6 through 11, a correlational analysis was conducted to determine
significant relationships between four independent latent variables.
Structural Equation Modeling
Overview
Structural Equation Modeling (SEM) was conducted by using LISREL 8.30.
LISREL 8.30 program defaults to the maximum likelihood fitting function for estimating
a model's parameters (Joreskog & Sorbom, 1996). This method consists of a
measurement model to define hypothetical latent constructs in terms of measured
variables, and a structural model to depict relationships among latent constructs. The
SEM is a multivariate method combining aspects of factor analysis and multiple
regression in analyzing a set of interrelated relationships among manifest and latent
variables simultaneously. The advantages of SEM compared to most other conventional
statistical methods include the following capabilities.
First, the basic statistic in SEM is the covariance. Covariance is generally a more
appropriate statistic in SEM than correlation, and covariance statistics convey more
information than correlations. A correlation matrix is not suggested since SEM assumes