Add missing data handling and allow GSCA correlation-only models

[Profalamer]

This PR adds two focused improvements:

1. Missing-data handling

• Adds automatic listwise deletion as the default behavior.
• Adds .missing = "mean" and .missing = "regression" options.
• Removes the previous stop-on-missing-data behavior.
• Records missing-data handling details in Information$Missing_data.
• Avoids repeated missing-data warnings during postestimation output.

2. GSCA correlation-only models

• Allows GSCA estimation when the model specifies construct correlations using ~~
  without structural paths using ~.
• This enables correlated composite models under GSCA.

Validation:

• devtools::test(filter = "gsca")
• Manual smoke checks for .missing = "listwise", "mean", and "regression".

[emstruong]

Hi @Profalamer ,

It's nice to meet you, thank you for your pull request.

I saw your earlier email and thought I might drop a note.

Currently, GSCA is undergoing a substantial revamp and it will take some time to reach to main. In the meantime, pull requests that affect GSCA code is a bit sensitive because it might affect the current (#594 and #600) and upcoming planned pull requests. The next planned pull request will be from the gscaBoot branch.

I'm not familiar with running GSCA models without a structural model, but my understanding is that you'd also have to change the way the alternating least squares algorithm handles the $B$ matrix. Have you considered this?

Also, just in case, some general contribution guidelines are listed here: https://github.com/FloSchuberth/cSEM/wiki/Contribution-Guidelines :)

Michael

[Profalamer]

Hi @emstruong

Thank you for your reply.

Running a GSCA with a correlated model should follow a similar logic to PLS with a correlated model. Please check this paper:
Schuberth F, Henseler J and Dijkstra TK (2018) Confirmatory Composite Analysis. Front. Psychol. 9:2541. doi: 10.3389/fpsyg.2018.02541

Unless I miss something, I don’t believe this should be a complex issue. At the same time, I understand that you and Florian are working on a formal paper revamp of some aspects of GSCA, and my proposal might clash with it. I’m still curious if the proposed change could be temporarily retained until the project is completed. You mentioned that the project might take some time, possibly even years, to be published.

And, yes, I'm aware of the general contribution guidelines.
Abdullah