Requirement for PLS-SEM

Questions about the implementation and application of the PLS-SEM method, that are not related to the usage of the SmartPLS software.
Post Reply
MeiPeng
PLS User
Posts: 15
Joined: Sun May 24, 2015 3:33 am
Real name and title: LowMeiPeng

Requirement for PLS-SEM

Post by MeiPeng »

Dear All

I just done my Work Completion Seminar (WCS) for my PhD studies.

I was being penalized for using PLS-SEM for my data analysis because I had collected 389 usable questionnaires and my data is normal.

I used and justified PLS-SEM to run my data based on the following reasons as extracted from Source: Adopted from Sarstedt, M., Ringle, C. M., Smith, D., Reams, R., & Hair, J. F. (2014). Partial least squares structural equation modeling (PLS-SEM): A useful tool for family business researchers. Journal of Family Business Strategy, 5(1), 105-115.:
1- My model comprises of 7 predictors, which is more than five and therefore, my model has many path relationships and many indicators per construct.
2- My model includes advanced model elements, 2 mediator variables , 1 moderator and 1 hierarchical component model, which is formatively measured constructs.

According to Hair, Hult, et al (2014) advised that it is important to verify that the data are not too far from normal distribution as extremely non-normal data prove problematic in the assessment of parameters’ significance.
Therefore, I run normality tests (Shapiro-Wilk test , Skewness and Kurtosis) and the results are not far from normal.

I was told that it is inappropriate for me to use PLS-SEM for the following reasons:
1- I have a large sample as it is above 100
2- My data shown to be normal.
3- I only have 1 construct which is formatively measured. I was told I need to have a min of 3 constructs that are formatively measured, then only I qualified to use PLS-SEM.

Can anyone help me here?

My questions are :
1- If my data is above 100 samples, I cant use PLS-SEM?
2- If my model only consist of 1 formatively measured construct, I cant use PLS-SEM?
3- If my data is normally distributed, I cant use PLS-SEM?

Appreciate all the experts reply and opinions.

Thanks and Regards
mei peng
janschreier
PLS Expert User
Posts: 116
Joined: Fri Sep 12, 2014 2:12 pm
Real name and title: Jan Schreier

Re: Requirement for PLS-SEM

Post by janschreier »

Hi Mei,

as a quick answer, what the feedback probably refers to is that you should use CB-SEM (LISREL or AMOS).

This site gives a hint when to use which:
https://www.smartpls.com/documentation/ ... and-cb-sem

As for the exact numbers ("more than 100") needed to calculate a certain model it depends on the model's complexity.

Hope that helps!
Jan
MeiPeng
PLS User
Posts: 15
Joined: Sun May 24, 2015 3:33 am
Real name and title: LowMeiPeng

Re: Requirement for PLS-SEM

Post by MeiPeng »

Dear Jan

Thank you for your swift reply.

I did the sample size calculation by using Gpower and I only need 103 samples. If I followed Cohen (1997), no of max arrow pointing to a construct, the sample size I need is 70.

Thanks and Best regards
mei peng
janschreier
PLS Expert User
Posts: 116
Joined: Fri Sep 12, 2014 2:12 pm
Real name and title: Jan Schreier

Re: Requirement for PLS-SEM

Post by janschreier »

Hi Mei,

I (and also the reviewer) do not complain about the fact that your sample size is too small but what your reviewer means is that: "your sample size is large enough to use CB-SEM but you did not use it (and instead used PLS)!".
Post Reply