I got a question about moderating effect, multigroup analysis with smartPLS. It is unclear to me that in each group need to ensure about measurement model criteria regarding reliability and validity (item loading, CR, Cronbach alpha, AVE). However, when I run separate by each sample group, there are not exactly the same number of items loads into the constructs. For example, Construct 1 of sample group 1 got 3 items but Construct 1 of sample group 2 got 5 items to explain the construct.
I am not sure if this is ok, but I look through a number of references but its not clear about there measurement model especially items loading in each sample group are the same.
If you can give me suggestions on this, it would be very helpful.