10 times rule

[Priyankafox]

Hi based on Hair Jr et al. (2013),

In PLS SEM, the sample size is determined by the often-cited 10 times rule which indicates that the sample size should be equal to the larger of:
a. 10 times the largest number of formative indicators used to measure a single construct, or
b. 10 times the largest number of structural paths directed at a particular construct in the structural model.

Since my model (attached to this post) uses all reflective constructs, I cannot use rule (a). Thus I use (b).

Based on the the diagram, in the context of my conceptual model, is the construct with the largest number of structural paths - brand image?

And does that mean the structural paths are 3?
So 30 is the minimum sample?

Sorry if this seems obvious. I am a novice and quite nervous. I would appreciate any feedback. Thanks.

Reference
Hair Jr, JF, Hult, GTM, Ringle, C & Sarstedt, M 2013, A primer on partial least squares structural equation modeling (PLS-SEM), Sage Publications.

[jmbecker]

Yes, based on that rule of thumb your minimum sample size is 30. However, that does not mean that such a sample size is sufficient for detecting small effects or for estimating robust parameters from which you can draw generalizable conclusions. A sample size of 30 is definitely not enough for such purposes.
It is better to conduct a power analysis with some assumptions about the parameters that you expect.

[Hengkov]

Hi,

Many researchers were mistaken in justifying the amount of sample. Generally, they are based on the minimum number of samples. In fact, the minimum sample size is not the best option. Why should use the minimum number of samples, if the number of samples can be collected is greater than that. The use of a large number of samples to be better.
By using G*Power: 3 predictors, effect size 0.15 and Sig 5% obtained the minimum number of samples = 77.

Best regards,

[Priyankafox]

Thanks so much Hengkov and jmbecker !!

I read up on effect size (f2) and tried to understand through past posts: viewtopic.php?f=2&t=461

If I want to calculate f2 in my model above.

First I get R2 of the whole model by running PLS algorithm.

To R2 excluded:
Then I delete one of the independent variables lets say risk perception. Then I run the PLS algorithm again and I get the R2 excluded.
I do this step for safety concern of industry and safety concern of industry.

Is this correct?


Also Q2 (predictive relevance)
I understand I have to use blindfolding.

However in my model I am slightly confused how to proceed.

When I am running blindfolding on SmartPLS, I am supposed to select the dependent construct but in my model I have
1. brand image and mediator
2. purchase intentions and repurchase intentions as dependent

What construct do i select for blindfolding? All 3 variables in 1 and 2?

I understand the steps are the following for Q2:
1. first get the q2 for the whole model by running blindfoling, choose omission distance, then look at cross validated redundancy
2. the get q2 excluded

However in step to get q2 excluded, when i run blindfolding, which construct do I select??

I get the theory but in practice I don't how to apply it on SmartPLS using my model. It would must easier if my model stopped at brand image but because I have brand image, purchase intentions and repurchase intentions- I am lost.

I would greatly appreciate some feedback.