Your browser is out-of-date!

Update your browser to view this website correctly. Update my browser now

×

Outer model weighting options

SmartPLS allows the user to specify four additional weighting modes for a construct. Thus, it allows advanced modeling options for constructs in your path model. Namely, Mode A, Mode B, Sumscores and Pre-Defined Weights.

Brief Description

SmartPLS allows the user to specify five different weighting options for a construct. Thus, it allows advanced modeling options for constructs in your path model. If you want to apply the standard PLS algorithm, you don't have to do anything, just leave this option on automatic (default). This is the same practice as known from previous versions of the SmartPLS software (e.g., SmartPLS 3.2.1 and lower).

If you use one of the other options you are able to model constructs based on the latest discussions around PLS path modeling. Note: You may not get standard PLS results. Many of the consequences of using the different options in various model settings have not been thoroughly tested, yet. Thus, only change the outer model weighting options if you are sure that you understand the consequences.

  • Automatic: (default) This is the standard option for all newly created variables. It uses the weighting Mode A and B (e.g., Lohmöller 1989; Wold 1982) according to the specified measurement of the construct. More specifically, the automatic option uses Mode A for reflective constructs and Mode B for formative constructs. This is the same practice as explained by Hair et al. (2014) and as known from previous versions of the SmartPLS software (e.g., SmartPLS 3.1; Ringle et al., 2015).
  • Mode A: Alternatively, users can specify a Mode A constructs. This mode uses the covariance between each indicator and the latent variable as weights for determining the latent variable scores. Based on these results and regardless of the measurement model specification (i.e., formative or reflective), one obtains the final outer weights, outer loadings and structural model relationships. Mode Acan be useful also for formative measurement models in some situations (e.g., for predictive models) as Becker, Rigdon and Rai (2013) show in detail.
  • Mode B: Users can also specify a Mode B constructs. This mode uses the coefficients of a multiple regression between the latent (i.e., dependent) variable and its indicators (i.e., independent variables) as weights for determining the latent variable scores. Based on these results and regardless of the measurement model specification (i.e., formative or reflective), one obtains the final outer weights, outer loadings and structural model relationships. For example, a reflective construct could use Mode B results estimation. However, this combination (i.e., reflective measurement model and Mode B) is not recommended and should only be used by experienced users for testing purposes.
  • Sumscores: Users can specify that a construct uses equal weights (unit weights), so that all indicators get the same weight when forming the latent variable scores. Some researchers recommend this option, especially when sample sizes are very small and estimation of weights is unstable. Otherwise, Sumscores usually provide inferior PLS-SEM results.
  • Pre Defined: Finally, users can specify that a constructs uses the pre-defined weights from the Initial Outer Weights dialog of the PLS algorithm. One can define any set of meaningful weights for the indicators of a construct. Usually, these weights stem from prior research or theory. When the user does not change anything in the Initial Outer Weights dialog, this option is equivalent to the Sumscores option, because all initial weight are equal.

References

Becker, J.-M., Rai, A., and Rigdon, E. E. 2013. Predictive Validity and Formative Measurement in Structural Equation Modeling: Embracing Practical Relevance. In 2013 Proceedings of the International Conference on Information Systems. Milan.

Hair, J. F., Hult, G. T. M., Ringle, C. M., and Sarstedt, M. 2014. A Primer on Partial Least Squares Structural Equation Modeling (PLS-SEM). Thousand Oaks, CA: Sage.

Lohmöller, J.-B. 1989. Latent Variable Path Modeling with Partial Least Squares. Heidelberg: Physica.

Ringle, C. M., Wende, S., and Becker, J.-M. 2015. "SmartPLS 3." Bönningstedt: SmartPLS GmbH.

Wold, H. O. A. 1982. "Soft Modeling: The Basic Design and Some Extensions." In Systems Under Indirect Observations: Part II. Eds. K. G. Jöreskog and H. O. A. Wold. Amsterdam: North-Holland, 1-54.