Composite, Common Factor and Mixed Models
Abstract
PLSSEM allows estimating proxies of latent variables that represent different model types (i.e., composite models and common factor models). One can use only latent variables of the one or the other type in a PLS path model. However, it also is possible to employ latent variables of both types in a PLS path model (i.e., mixed models).
Composite Models
PLS principally estimates composite models when using the PLS algorithm. When a reflective measurement model is considered (i.e., with relationships from the construct to the indicators), the PLS algorithm computes the composites using Mode A (i.e., the outer weights are the correlations between the construct and the indicators). In case of a formative measurement model (i.e., with relationships from the indicators to the construct), the PLS algorithm computes the composites using Mode B (i.e., the outer weights are the multiple regression coefficients with the indicators as independent variables and the latent variable as dependent variable). By double clicking on the latent variable, one can change the computation method of the composites (e.g. to sumscores).
Common Factor Models
When using PLSc, it is possible to mimic common factor model results. When a reflective measurement model is considered (i.e., with relationships from the construct to the indicators), the PLSc algorithm computes the composites using Mode A. Then, the composite’s relationships in the measurement model and to other latent variables in the structural model are corrected for attenuation and, as a results, mimic the common factor model results.
Mixed Composite and Common Factor Models
When using PLSc, it is possible to mimic common factor model results. When a reflective measurement model is considered (i.e., with relationships from the construct to the indicators), the PLSs algorithm computes the composites using Mode A . Then, the composite’s relationships in the measurement model and to other latent variables in the structural model are corrected for attenuation and, as a results, mimic the common factor model results. In case of a formative measurement model (i.e., with relationships from the indicators to the construct), the PLSc correction is not applied. The PLS algorithm computes the composites using Mode B. By double clicking on the latent variable, one can change the computation method of the composites (e.g., switch it from Mode B to Mode A).
Switching Modes in SmartPLS
Per default, SmartPLS 3 uses Mode A (correlations weights) estimations for reflective measurement models and Mode B (regression weights) for formative measurement models. When doubleclicking on a construct, you can specify a specific estimation (Mode A, Mode B, sumscores, predefined weights) of the construct.
Links
References

Hair, J. F., Hult, G. T. M., Ringle, C. M., Sarstedt, M., and Thiele, K. O. (2017). Mirror, Mirror on the Wall: A Comparative Evaluation of Compositebased Structural Equation Modeling Methods, Journal of the Academy of Marketing Science, in press.

Henseler, J., Dijkstra, T. K., Sarstedt, M., Ringle, C. M., Diamantopoulos, A., Straub, D. W., Ketchen, D. J., Hair, J. F., Hult, G. T. M., and Calantone, R. J. (2014). Common Beliefs and Reality about Partial Least Squares: Comments on Rönkkö & Evermann (2013), Organizational Research Methods, 17(2): 182209.
 Sarstedt, M., Hair, J. F., Ringle, C. M., Thiele, K. O., and Gudergan, S. P. (2016). Estimation Issues with PLS and CBSEM: Where the Bias Lies! Journal of Business Research, 69(10): 39984010.