Using Partial Least Squares in Digital Government Research

Using Partial Least Squares in Digital Government Research

J. Ramon Gil-Garcia (Centro de Investigación y Docencia Económicas, Mexico)
Copyright: © 2008 |Pages: 15
DOI: 10.4018/978-1-59904-857-4.ch023
OnDemand PDF Download:
$37.50

Abstract

Digital government is a complex socio-technical phenomenon, which is affected by technical, managerial, institutional, and environmental factors (Dawes & Pardo, 2002; Fountain, 2001; Gant, 2003; Garson, 2000; Heeks, 2005; Kraemer, King, Dunkle et al., 1989; Landsbergen & Wolken, 2001; Laudon, 1985; Rocheleau, 2003). Recent studies have greatly contributed to developing the necessary knowledge about e-government benefits and success factors (Barrett & Greene, 2000; Dawes, 1996; Gil-García & Pardo, 2005; Heeks, 2003; Holmes, 2001; O’Looney, 2002; Rocheleau, 1999; West, 2005; Zhang et al., 2002). However, an important portion of this research has used a single measure of e-government and relatively simple assumptions about the relationships between information technologies and organizational, institutional, and contextual factors (Gil-García, 2005b).

Key Terms in this Chapter

Indicator: It is a variable that can be directly observed and measured and, in a SEM framework, it is used to form or reflect a construct or latent variable.

Construct: It is a variable that cannot be observed and measured directly, but has theoretical relevance. In a SEM framework, multiple indicators are used to form or reflect a construct or latent variable.

Partial Least Squares (PLS): SEM technique that attempts to minimize the differences between the observed and predicted values of endogenous constructs (dependent variables).

Indirect Effect: In a causal model, this is the impact of one variable on another, through the impact of the former on a third variable that is called mediating variable. In complex models, several indirect-effect paths may exist between two variables.

Measurement Error: This type of error is derived from the quality of the indicators used to operationalize a variable. SEM techniques allow researchers to model the measurement error, while other statistical techniques, such as linear regression, assume no measurement error.

Direct Effect: In a causal model, this is the direct impact of one variable on another and is represented by an arrow from the independent to the dependent variable.

Covariance-Based SEM: SEM techniques that attempt to minimize the differences between the observed and predicted covariance matrices.

Structural Equation Modeling (SEM): Statistical technique that allows simultaneously testing the measurement and structural models. Two of the main types of SEM are covariance-based and variance-based.6

Complete Chapter List

Search this Book:
Reset