Folks,
I would like to evaluate discriminant validity of my 2nd order reflective-formative model.
I am not 100% sure if this is the right way to do it:
Employing the two stage-approach I first created my model using the repeated indicator
approach, ran the PLS algorithm and saved the latent variables in a .csv file. Based on this
data I built my "new" model and now would use the HTMT ratios as well as the confidence
interval to evaluate discriminant validity.
However, my new model does not include my lower-order-constructs, but rather only my
higher order construct.
Would this still be the right approach?
Best regards
Chris
Reflective-Formative HCM | Discriminant Validity HTMT
-
- PLS User
- Posts: 16
- Joined: Thu Oct 17, 2013 10:04 am
- Real name and title: Marc Janka
- Location: Germany
Re: Reflective-Formative HCM | Discriminant Validity HTMT
Hi Chris:
I would advise you not to use the repeated indicator approach because of issues that arise of artificially correlated residuals (see Becker et al., 2012). Instead I would use the two step approach and first estimate the latent variable scores of your first-order construct in a fully specified model without the second-order construct. Then I would use these latent variable scores in the second step.
For formative constructs, construct validity and discriminant tests are not feasible. Instead, formative constructs (and this also concerns second-order constructs) require in-depth theoretical explanation (see Sarstedt et al., 2016) and you should evaluate if multicollinearity may be an issue.
Becker, J.-M., Klein, K., and Wetzels, M. 2012. Hierarchical Latent Variable Models in PLS-SEM: Guidelines for Using Reflective-Formative Type Models. Long Range Planning 45 (5–6): 359–394.
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): 3998−4010.
Best
Marc
I would advise you not to use the repeated indicator approach because of issues that arise of artificially correlated residuals (see Becker et al., 2012). Instead I would use the two step approach and first estimate the latent variable scores of your first-order construct in a fully specified model without the second-order construct. Then I would use these latent variable scores in the second step.
For formative constructs, construct validity and discriminant tests are not feasible. Instead, formative constructs (and this also concerns second-order constructs) require in-depth theoretical explanation (see Sarstedt et al., 2016) and you should evaluate if multicollinearity may be an issue.
Becker, J.-M., Klein, K., and Wetzels, M. 2012. Hierarchical Latent Variable Models in PLS-SEM: Guidelines for Using Reflective-Formative Type Models. Long Range Planning 45 (5–6): 359–394.
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): 3998−4010.
Best
Marc