Clarification on Full Mediation in Structural Model

[Yashi Kapil]

Respected group members and Neeraj Sir

I would like to seek clarification regarding the interpretation of mediation in my research model. The analysis has been conducted using SmartPLS (PLS-SEM). In my results, I observed that the independent variables do not have a significant direct effect on the dependent variable. However, they show a significant effect on the dependent variable through the mediator.

Based on this, I understand that this is a case of full mediation. I would like to confirm whether such a model where both the IVs show full mediation can be considered theoretically and statistically valid.

Additionally, I would appreciate your guidance on how best to justify and report this finding in my research.

Looking forward to your valuable insights.

Thanks, and regards

Yashi Kapil

Research Scholar, Bennett University

[Dr. Aman Kumar]

Yes, Kapil  this can be considered Valid ......but check that IDV----->DV and IDV--->Mediator----->DV there must be a theory of their relationship in the literature.

--
The members of this group are expected to follow the following Protocols:
1. Please search previous posts in the group before posting the question.
2. Don't write the query in someone's post. Always use the option of New topic for the new question. You can do this by writing to dataanaly...@googlegroups.com
3. It’s better to give a proper subject to your post/query. It'll help others while searching.
4. Never write Open-ended queries. This group intends to help research scholars, NOT TO WORK FOR THEM.
5. Never write words like URGENT in your posts. People will help when they are free.
6. Never upload any information about National Seminars/Conferences. Send such information
in personal emails and feel free to share any RESEARCH-related information.
7. No Happy New Year, Happy Diwali, Happy Holi, Happy Birthday, Happy Anniversary, etc. allowed in this group.
8. Asking or sharing Research Papers is NOT ALLOWED.
9. You can share your questionnaire only once.
---
You received this message because you are subscribed to the Google Groups "DataAnalysis" group.
To unsubscribe from this group and stop receiving emails from it, send an email to dataanalysistrai...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/dataanalysistraining/PN3PPF5F2586B8BBBA60EA716A21B67A4ACF450A%40PN3PPF5F2586B8B.INDPRD01.PROD.OUTLOOK.COM.

--

Aman Kumar

[Muhammad R Siregar]

Dear Yashi,

Full mediation is theoretically and statistically justified, albeit rarely found in practice. I am not saying that your conclusion is wrong—let us be clear about it. However, if I were you, I would check whether these two artifacts exist in the study: (1) low statistical power; and (2) reverse causality.

(1) Low statistical power

True effect may be too small to detect with your current sample size. For practical heuristics, you may check bootstrap CI for direct path. If it is wide and includes zero but also large positive/negative values (e.g., β = 0.05, 95% CI [‑0.40, 0.50]), that usually tells that your study is underpowered. Your data cannot reliably tell whether the direct effect is zero, moderately positive, or moderately negative. Wide interval may also occur, even with large sample, when the direct effect is too variable (e.g., with high measurement error), but generally it indicates imprecise estimates.

To really establish that you have enough power, you can run post-hoc power analysis for the smallest direct effect you consider theoretically meaningful (e.g., β = 0.10) and compare your sample with the recommended sample size for that effect. If your sample is below the recommended size, your study is indeed underpowered.

(2) Reverse causality

Reverse causality could theoretically explain full mediation conclusion. In cross‑sectional data, it is possible that Y actually causes M (rather than M causing Y, and therefore your model becomes misspecified). If that were true, the algorithm would force some X-Y covariance through M, producing significant indirect effect—that would be spurious mediation (the path exists statistically but not causally). For the direct path, M becomes the collider. Statistically, controlling for a variable that is caused by the outcome (Y) may induce a non-significant direct effect.

You cannot check for reverse causality, though. But you can provide theoretical justification that X precedes M, and M precedes Y. If the literature were to suggest that Y influences M, it would be best to acknowledge and report it as a limitation.

I wish you the best of luck with your research.

Regards,

MRS

To view this discussion visit https://groups.google.com/d/msgid/dataanalysistraining/CAFBvfJpL4ZJ8kZJzeFbZ7tWptG24CZVWQy%2BvwempZRRyH%3DA1Tg%40mail.gmail.com.

[Yashi Kapil]

Thank you so much for the insights, sir. This is helpful for my research. 

Regards

Yashi Kapil.

Get Outlook for iOS

---

From: dataanalys...@googlegroups.com <dataanalys...@googlegroups.com> on behalf of Muhammad R Siregar <sire...@gmail.com>
Sent: Friday, April 3, 2026 8:09:15 PM
To: dataanalys...@googlegroups.com <dataanalys...@googlegroups.com>
Subject: Re: Clarification on Full Mediation in Structural Model

 

To view this discussion visit https://groups.google.com/d/msgid/dataanalysistraining/CAEG%3D6i1HHH%3DxWyVo-a2TymRB0GHBxs4S8j8f8ZmfH9i0HVCm0A%40mail.gmail.com.