Dear all,
I'm working in a field which hasn't been explored theoretically and empirically very much. My study is therefore of an explorative nature. I've read that this is generally possible/acceptable in PLS.
I know that my sample size will be quite small (as I'm looking at a small industry with only few players). At the same time, in my PLS-model I want to test quite a lot of potential LVs and correlations between them. I will thus most likely not be able to reach a sample size which satisfies the rule of thumb that n should be at least 10 times as large as the endogenous LV with the highest number of structural paths directed towards it.
My questions:
1) How can I handle this situation in PLS? Do I run PLS with all LVs included (violating the rule of thumb) and then throw out the insignificant structural paths, hoping that this will leave me with so few paths that the rule of thumb will be met in the end? Or would I rather take an iterative approach, where I test only a part of the model first (adhering to the rule of thumb), get rid of insignificant structural paths, and then add new LVs to the model?
2) I read on Chin's website that as a weak rule of thumb a sample size 5 times as large as the endogenous LV with the highest number of structural paths directed towards it would be satisfactory. Does anyone have a 'proper' literature source for this?
Thank you very much in advance!
Anke