insignificant weight

[torskiy]

I have a model that use both reflective and formative.
for example, the model is F--> R
F have dimension F1, F2, F3 & F4

formative indicator:

• F1a, F1b, F1c --> F1

reflective indicator:

• F2a, F2b, F2c <-- F2
  F3a, F3b, F3c <-- F3
  F4a, F4b, F4c <-- F4
  R3a, R3b, R3c <-- R3

1. Outer Model Test:

• Formative indicator shows that VIF < 5; but the significance weight vary
  • F1a = 8,017
    F1b = 0,936
    F1c = 4,068
  What should i do since the significance weight of F1b < 1,96 and what does it mean? can i keep it and run the next step?

2. Hypothesis: Fpositively influence R

• The Result for path coefficient shows:
  F1 --> R ; -1,000 ; t-stat insignificant
  F2 --> R ; 0,3670 ; t-stat significant
  F3 --> R ; -0,015 ; t-stat insignificant
  F4 --> R ; 0,1380 ; t-stat insignificant

Can i accept the hypothesis?

[jmbecker]

1)
You can find guidelines on formative model evaluation in the PLS-SEM book: https://www.smartpls.com/documentation/ ... s-sem-book
An insignificant weight is not a problem per se. The weight represents the relative contribution, given the other indicators, which could be insignificant. Yet, you should also check the loading (the absolute contribution/correlation with the construct). If the loading is significant and you can retain the indicator and just report that the weight is insignificant. If the loading is also not significant, you should consider removal of the indicator.

2)
You can only accept those hypothesis where the effect is significant.