Links to specific topics

Wednesday, September 15, 2010

There is no need for two-way arrows in WarpPLS


In covariance-based structural equation modeling (SEM) software tools, often one has to explicitly model correlations between predictor latent variables (LVs) to obtain related estimates. In WarpPLS, correlations between predictor LVs are automatically taken into consideration in the calculation of path coefficients.

That is, the path coefficients calculated by WarpPLS are true standardized partial regression coefficients, of the same type as those calculated through multiple regression analysis. The difference is, of course, that in WarpPLS the model variables are LVs, which are usually measured through more than one indicator. With multiple regression, only one measure (or indicator) is used for each variable in the model.

1 comment:

Ned Kock said...

I should also note that ALL correlations among latent variables are reported by WarpPLS, under the “View correlations among latent variables” option, after Step 5 is completed.