Hello there,
I am so grateful to have found this forum via a Google search as I have a question regarding sample size and the use of Smart-PLS.
I recently had my PhD viva and within the viva, one of my examiners questioned my use of Smart-PLS for my sample size and type of data (noting a strong preference for a CB-SEM instead). The examiner stated that Smart-PLS should ideally be used with smaller sample sizes and non-parametric data. My sample size was around 370 and my questionnaire was comprised of Likert scales (which is non-parametric data as it is ordinal). Does anyone have any references that support the use of Smart-PLS for larger samples sizes and specific rationale for why it is optimal with Likert scales? I have been asked to better justify its use in my corrections (I did state that within my discipline, Marketing, there has been a growing trend within the literature of the use of Smart-PLS and clearly outlined the difference between CB and PLS SEM within my thesis).
Thank you in advance.
Smart PLS and Sample Size
-
- SmartPLS Developer
- Posts: 1284
- Joined: Tue Mar 28, 2006 11:09 am
- Real name and title: Dr. Jan-Michael Becker
Re: Smart PLS and Sample Size
Better arguments than sample size are usually based on:
1) complex model relative to sample size. PLS models can be much more complex with limited sample sizes. Some CB-SEM models might not fit, even with sample sizes of 370, if the model is very complex.
2) formative measurement models. If you use formative measures, especially in endogenous positions, then PLS is clearly advisable. If you use only reflective indicators, CB-SEM might indeed be better.
3) predictive abilities. PLS allows the calculation of latent variable scores that conform to the model. CB-SEM does not allow this, because of factor indeterminacy. If you want to predict cases, for example, in a hold-out-prediction task, PLS is clearly advisable. Another advantage here might be, that you want to show the latent variable scores relative to the impact of the variables (in an IPMA analysis).
You find other advantages is the literature, but they have been in the center of some debates in the past.
1) complex model relative to sample size. PLS models can be much more complex with limited sample sizes. Some CB-SEM models might not fit, even with sample sizes of 370, if the model is very complex.
2) formative measurement models. If you use formative measures, especially in endogenous positions, then PLS is clearly advisable. If you use only reflective indicators, CB-SEM might indeed be better.
3) predictive abilities. PLS allows the calculation of latent variable scores that conform to the model. CB-SEM does not allow this, because of factor indeterminacy. If you want to predict cases, for example, in a hold-out-prediction task, PLS is clearly advisable. Another advantage here might be, that you want to show the latent variable scores relative to the impact of the variables (in an IPMA analysis).
You find other advantages is the literature, but they have been in the center of some debates in the past.
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
Researchgate: https://www.researchgate.net/profile/Jan_Michael_Becker
GoogleScholar: http://scholar.google.de/citations?user ... AAAJ&hl=de
-
- PLS Junior User
- Posts: 4
- Joined: Mon Jan 16, 2017 4:40 pm
- Real name and title: JB - PhD Student
Re: Smart PLS and Sample Size
jmbecker,
Thank you so much for taking the time to reply. All three points make sense.
I very much appreciate it.
Kind regards,
PBAUSA
Thank you so much for taking the time to reply. All three points make sense.
I very much appreciate it.
Kind regards,
PBAUSA