Morning!

If your model is reflective, you can try to previously rum model fit tests in AMOS and then run the SmartPLS. I don't know how many relationships you are testing, but the sample size is dependent on them. UTAUT model (if I remember correctly)tests about 7 relationships. Adding a moderator to all of them will produce about 7 additional relationships (7 moderations and one direct effect). The rule of thumb is about 30 participants for each relationship. So, you might be below the cut-off. But if model fit tests are fine, you can just use them as a justification. Look for our AOM paper from last year on Dynamic capabilities in highly dynamic environments. We used AMOS, but we had a small sample size. We also cited another A-Journal paper which handled this issue using model fit statistics (RMSEA, Chmin/DF, CFI, GFI, PClose).

Best,

Eugene

You will find a number of different "rules of thumb" for how many observations you need for each parameter you estimate (I think 30 is high; I see 10 mentioned more often).

In any event, running a moderated analysis with two groups will double the number of parameters you estimate, and thus double the size of the desired sample.

Hi Yevgen and David! 

Thanks very much for your responses. If I split my data into two groups I will probably use the model fit as justification, despite the lower sample size. 

I'm also using Smartpls, and it is quite well cited that you have greater flexibility when it comes to sample size using PLS SEM.