Statistics Glossary → Inner weighting scheme (PLS-SEM)
Inner weighting scheme (PLS-SEM)concept
The choice, in the PLS path modelling algorithm, of the rule by which each latent construct's inner proxy is formed from its neighbouring constructs.
The PLS-SEM algorithm is iterative, with two steps in each round. In the outer step each construct's score is computed as a weighted sum of its own indicators. In the inner step an inner proxy is formed for each construct from the scores of the constructs it is linked to in the structural model, and the inner weighting scheme sets the weights of that proxy. The path scheme uses multiple regression coefficients for a construct's predecessors and correlations for its successors. The centroid scheme (Wold) uses only the sign of the correlations (+1 or −1), and the factor scheme (Lohmöller) uses the correlations themselves.
The path scheme is the only one that takes the directions of the structural model into account and is usually recommended as the default. The centroid scheme can cause problems when correlations are close to zero, because their signs become unstable. In practice the three schemes often give very similar results, so if the choice changes the results noticeably, this should be examined as a sign of model instability. The scheme used should be reported.
An everyday example
A model proposes that 'service quality' and 'price perception' affect 'customer satisfaction', and that satisfaction affects 'loyalty'. Under the path scheme, the inner proxy of satisfaction is formed by multiple regression from its predecessors, quality and price, and by correlation from its successor, loyalty. Under the factor scheme all four links are weighted by correlations.
In MerQur
In PLS-SEM, the 'Inner weighting scheme' option chooses between the Path (default), Centroid and Factor (factorial) schemes.
Analyses where this concept appears
PLS-SEM (Partial Least Squares Path Modeling)