Interpreting Path Coefficients

Questions about the implementation and application of the PLS-SEM method, that are not related to the usage of the SmartPLS software.
Post Reply
rjb285
PLS Junior User
Posts: 9
Joined: Mon May 22, 2017 12:32 am
Real name and title: Ryan

Interpreting Path Coefficients

Post by rjb285 »

I have two questions in regard to interpreting path coefficients from bootstrapping. This is mainly interpreting the Original Sample (O), as I'm assuming that the Sample Mean (M) does not need to be interpreted, correct?

1. Is the path coefficient, Original Sample (O), interpreted as expressing the size of a relationship between two latent constructs (e.g., X has the largest, positive relationship with Y [O = 0.45; SD = 0.11]) or the size of the effect between two latent constructs (e.g., X has the largest, positive indirect effect on Y [O = 0.45; SD = 0.11])?

2. Would you mind providing a reputable citation for interpreting the Original Sample (O) path coefficient, rather than the Sample Mean (M) path coefficient, if it would be needed/helpful to do so?

Thank you.
jmbecker
SmartPLS Developer
Posts: 1282
Joined: Tue Mar 28, 2006 11:09 am
Real name and title: Dr. Jan-Michael Becker

Re: Interpreting Path Coefficients

Post by jmbecker »

The original sample coefficient is the one that you would also get from your normal PLS analysis and is the one that is estimated based on the sample provided. You should interpret this coefficient --> correct.
The sample mean in bootstrapping is the average coefficient over all bootstrapping runs. It indicates wether there exists some bias between original sample coefficient and sampling distribution. If the bias is large it is better to use bias-corrected confidence intervals for assessing the significance of the relationship.

The PLS-SEM book would be a possible reference: https://www.smartpls.com/documentation/ ... s-sem-book

You find direct, indirect and total effects in you bootstrapping output.
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
rjb285
PLS Junior User
Posts: 9
Joined: Mon May 22, 2017 12:32 am
Real name and title: Ryan

Re: Interpreting Path Coefficients

Post by rjb285 »

Thanks a lot. This one aspect is still left unclear (i.e. how to interpret the path coefficient):

1. Is the path coefficient interpreted as expressing the size of a relationship between two latent constructs (e.g., X has the largest, positive relationship with Y) or the size of the effect between two latent constructs (e.g., X has the largest, positive indirect effect on Y?

Thanks again. All of your help is really appreciated.
jmbecker
SmartPLS Developer
Posts: 1282
Joined: Tue Mar 28, 2006 11:09 am
Real name and title: Dr. Jan-Michael Becker

Re: Interpreting Path Coefficients

Post by jmbecker »

The path coefficient is interpreted like a standardized regression coefficient.
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
rjb285
PLS Junior User
Posts: 9
Joined: Mon May 22, 2017 12:32 am
Real name and title: Ryan

Re: Interpreting Path Coefficients

Post by rjb285 »

I'm sorry, to clarify, does this mean that it is interpreted as an R^2 value (coefficient of determination)?
jmbecker
SmartPLS Developer
Posts: 1282
Joined: Tue Mar 28, 2006 11:09 am
Real name and title: Dr. Jan-Michael Becker

Re: Interpreting Path Coefficients

Post by jmbecker »

no.
A path coefficient is interpreted: If X changes by one standard deviation Y changes by b standard deviations (with b beeing the path coefficient).
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
rjb285
PLS Junior User
Posts: 9
Joined: Mon May 22, 2017 12:32 am
Real name and title: Ryan

Re: Interpreting Path Coefficients

Post by rjb285 »

I'm sorry, to clarify once again, how is the layman going to understand this, though, when reading an article? For example, can the increase in 1 unit SD be interpreted as a sizable effect and can this effect be quantified (e.g., small, medium, large)?

Thanks again, and for your patience.
mklacmer
PLS Junior User
Posts: 3
Joined: Fri Aug 30, 2019 2:01 pm
Real name and title: Mario Klačmer

Re: Interpreting Path Coefficients

Post by mklacmer »

Im reading one article and the author say that "path coefficient (beta) is 0,446 and so we can say that perceived control had a strong effect on intentions to use".
Some other betas he describes as moderate and some as small.
But how he conclude that?
What are the references values for path coefficients in PLS-SEM?
I found rules of thumb for R2 and Q2, but no for path coefficients.
Many thanks!
jmbecker
SmartPLS Developer
Posts: 1282
Joined: Tue Mar 28, 2006 11:09 am
Real name and title: Dr. Jan-Michael Becker

Re: Interpreting Path Coefficients

Post by jmbecker »

The normal PLS path coefficients are interpreted like standardized regression coefficients. Thus, they can be descriptively compared in their magnitude because they are all on the same scale. However, to make justified claims about one being larger than the other you may also want to test this using the approach from the following paper:
Rodríguez-Entrena, M., Schuberth, F., & Gelhard, C. (2018). Assessing statistical differences between parameters estimates in Partial Least Squares path modeling. Quality & Quantity, 52(1), 57-69.
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
Tonka
PLS Junior User
Posts: 8
Joined: Mon Oct 19, 2020 2:47 pm
Real name and title: Antonija Petrlić, MBA

Re: Interpreting Path Coefficients

Post by Tonka »

Dear,

please just for advice ie. clarification. When defining hypothesis in many research I see the use of different phrases "positive effect", "positive impact", "positive influence" and "positive relationship". What is the correct phrase to use?
Is it correct to say that "variable x has a positive impact on y" or "variable x has a positive relationship with y"?
Thank you very much.
Kind regards,
Antonija
Post Reply