Hi Flo,

So just a part of the series of Qs for our eventual online meeting (meaning not currently urgent AFAICT):

Currently, cSEM generally estimates the structural model independently of the weights and loadings. That's fine, although this seems awkward when harmonizing with GSCA-type methods because GSCA naturally obtains the path coefficients as part of its ALS routine (using OLS).

The primarily relevant file in question is: estimators_paths.R

estimatePath() computes a number of important statistics outside of structural coefficients, so it probably shouldn't just be side-stepped.

I'm not sure if this can just be transplanted atop the GSCA routines and if it might be making assumptions that are not compatible with GSCA...

Of the series of issues, this is the one that's probably most likely to simply 'OK'. Honestly, I should probably step through the function myself slowly to see what it's really doing... But anyways, this can be left for fine-tuning.

I guess the objective is to verify whether the following statistics as computed via estimatePath() are applicable to the context of GSCA:

• Path Coefficients
• R2
• R2adj
• VIF
• SE

Edit: Actually, this procedure is probably incorrect for GSCA-M because it uses the wrong construct scores $XW$, as opposed to $(X-UD)W$.