Bootstrap CI of standardized coefficients in mediation analysis and Hayes' recommendation

Questions about the implementation and application of the PLS-SEM method, that are not related to the usage of the SmartPLS software.
Post Reply
lunar0413
PLS Junior User
Posts: 2
Joined: Fri Oct 11, 2019 4:43 pm
Real name and title: David

Bootstrap CI of standardized coefficients in mediation analysis and Hayes' recommendation

Post by lunar0413 »

Hello,

I am new to SmartPLS3 and trying to use PLS-SEM for my research for the first time.

While I was working on my model, I had a question about Bootstrap CI of the standardized coefficients of indirect effects. According to Hayes, Bootstrap CIs of standardized coefficients in mediation analysis (indirect effects) have to be interpreted with caution because the bootstrapping procedure does not properly generate the confidence interval for standardized coefficients. I believe there are some published papers (e.g., Cheung, 2009, Preacher & Selig, 2012) and there is a discussion about this topic in FAQs of Hayes' website (http://processmacro.org/faq.html). I believe this is one of the reasons why Hayes recommends to use unstandardized coefficients bootCI and Hayes' PROCESS provides a separate feature that researchers can get a proper confidence interval (i.e., effect size option).

So, here are my questions.
1) Does the bootstrap CI of the standardized coefficient of indirect effects in SmartPLS3 consider Hayes' suggestion? In other words, is it ok to interpret the bootstrap CI for the standardized coefficient of indirect effects as a proper (correct) confidence interval? I believe Hayes suggested that to get a proper bootCI for standardized coefficients, all the bootstrapped samples have to be separately standardized in each bootstrapping process. For example, if I use 5,000 times of iteration, 5,000 samples have to be separately standardized (Cheung, 2009, p. 427).

2) I found that I can get the unstandardized coefficients for indirect effects in IPMA results. Are there any ways that I can get the bootCI for these unstandardized coefficients?

I would like to make sure that the results from SmartPLS3 account for the issue with BootCI for standardized coefficients. If not, I would like to report unstandardized coefficients and their unstandardized boot CI along with standardized coefficients.

Thank you. I look forward to hearing from someone has knowledge on this matter or the developer of SmartPLS3.
jmbecker
SmartPLS Developer
Posts: 1282
Joined: Tue Mar 28, 2006 11:09 am
Real name and title: Dr. Jan-Michael Becker

Re: Bootstrap CI of standardized coefficients in mediation analysis and Hayes' recommendation

Post by jmbecker »

1) Yes. The bootstrapping procedure standardizes each sample and not the original data that it samples from. In this sense we are following Hayes recommendation, although this has been the standard practice in PLS much longer than Hayes Macros exist. Because it is something not only important for the indirect effect, but also for the normal (direct) path coefficients and all other parameters that are all standardized effects in the basic PLS.

2) Currently not. We are considering to offer a bootstrap of the IPMA results, but this is something for future versions. At the moment demand for such a feature seems very low and our priorities are elsewhere. Your request is just the second that we received.

We are currently working on SmartPLS 4, where we are redesigning some core modelling features and also want to allow more flexibility in modelling (mimicking) PROCESS models.
Dr. Jan-Michael Becker, BI Norwegian Business School, SmartPLS Developer
Researchgate: https://www.researchgate.net/profile/Jan_Michael_Becker
GoogleScholar: http://scholar.google.de/citations?user ... AAAJ&hl=de
lunar0413
PLS Junior User
Posts: 2
Joined: Fri Oct 11, 2019 4:43 pm
Real name and title: David

Re: Bootstrap CI of standardized coefficients in mediation analysis and Hayes' recommendation

Post by lunar0413 »

Dr. Becker,

Thank you so much for your quick response. I am looking forward to the new version of SmartPLS.
Post Reply