Can you account for nested data (individuals within organisations) in PLS-SEM?

[Paul Freeman]

Hi,

Is it possible to account for nested data within PLS-SEM (e.g., can you run multi-level models)? If so, can anyone offer any suggestions on appropriate software programs, documents explaining how this can be done, and/or examples of articles that have done this?

To elaborate a little, I hope to collect data from individuals nested/clustered within organizations. The primary construct of interest has 4 dimensions each assessed with 5-15 formative indicators. One purpose of the research would be to examine the proposed structure of this construct/questionnaire (i.e., assess the outer measurement model). A second purpose would be to examine the relationships between this construct and three outcomes of interest (each is reflective, unidimensional and assessed by 5-10 indicators). Throughout my analyses, I would like to account for potential non-independence in the data if possible – that respondents are nested within organizations.

Any advice would be very much appreciated.

 

Paul

[Ned Kock]

You can do this type of analysis via a growth curve analysis employing the current nonlinear features of WarpPLS. I am planning on writing a post on the WarpPLS blog about this  - hopefully soon. In the meantime, I suggest you check the resources available from warppls.com.

 

Best regards, Ned

---

This email is free from viruses and malware because avast! Antivirus protection is active.

[Kai]

Hi Ned,

I am sure many of the warppls users are looking forward to your post on this topic, I am!

Could you please tell us for the time being which of the online resources you refer to exactly? And would you happen to know a good exemplary paper that already used this approach (with warppls)?

Thanks and best regards,
Kai

[Ned Kock]

Hi Kai.

 

> I am sure many of the warppls users are looking forward to your post on this topic, I am!

 

It is not a trivial issue, involving combined linear-moderating and nonlinear analyses with different model configurations, which is why the post is still in the works.

 

Ned