Component-based Path Modeling Open Issues and Methodological Contributions

In this work we discussed some issues in PLS-PM and proposed methodological contributions to enhance PLS-PM potentialities. PLS-PM is a component-based method for SEM. Instead of severing every tie between component-based methods and factor-based methods we think that researchers should commit themselves in finding out which approach works best in which circumstances, and a continuous dialogue between the two communities of researchers is highly recommended for progress in this area of research. We compared PLS-PM and ML-SEM in the framework of the same simulation design, investigating the effects of measurement model misspecification and the implications of formative MVs on both methods. A common impression found in the literature is that only PLS-PM allows the estimation of SEM including formative blocks. The implication of formative MVs in Covariance-Based framework is a rather difficult task. However, if certain model specification conditions are satisfied the model is identified, and it is possible to estimate a Covariance-Based SEM with formative blocks (Bollen and Davis, 2009; Williams et al., 2003). Measurement model misspecification has the potential for poor parameter estimates and misleading conclusions (see Dolce and Lauro, 2014; Jarvis et al., 2003; MacKenzie et al., 2005, among others). Its effects extend also to the estimates of the path coefficients connected to the misspecified block. Our simulation results showed that the effect of measurement model misspecification is much larger on the ML-SEM parameter estimates. Besides considering PLS-PM as an alternative method for SEM, PLS-PM is a descriptive and prediction oriented method, deserving a prominent place in research applications when the aims of the analysis is prediction (Becker et al., 2013). For this reasons, further studies on the predictive ability of PLS-PM are needed. The PLS-PM evaluation criteria should include the predictive ability and further criteria and evaluation techniques for PLS-PM are needed (Sarstedt et al., 2014). Based on the proposed criteria, further extensions and modifications should be made on the basic PLS-PM algorithm in order to improve the predictive capabilities of the model estimation. The non-symmetrical approach for component-based path modelling (NSC-PM) presented in the third chapter of this dissertation is an example of work in this direction. NSC-PM is a non-symmetrical approach that aims at maximizing the explained variance of the MVs of the endogenous and bridge blocks ( i.e. an approach based on the optimization of a redundancy-related criterion in a multi-block framework). Unlike PLS-PM, NSC-PM respects the direction of the relationship specified in the Path diagram (i.e. the path directions), since the directions of the links in the inner model play a role in the algorithm. Compared to the other component-based methods, NSC-PM seems to be a good compromise between favoring stability (high explained variance) in the blocks and correlation between components. In the last chapter of the thesis we presented the Quantile Composite-based Path Modelling (QC-PM). QC-PM exploits both Quantile regression (QR) (Koenker and Basset, 1978) and quantile correlation (QC) (Li et al., 2014), which allow respectively the estimation of a set of conditional quantile functions and a correlation measure to examine the linear linear relationships between any two variables for different quantiles, providing a more complete picture of the relationships between varia