How assess, report and interpret Moda A weights

[DetHeider]

Hello gentlemen,

I saw some suggestions to calculate weights for formative indicators with Mode A -> correlation weights.
When I do this do I have to evaluate the results the same way like when I would have used Mode B? (VIF, weight size, weight significance, loading size, loading significance -> not significant or not high -> delete indicator)? And how should the results be reported? with loadings, weights, significance levels?
And the weights can be interpreted the same way as when using Mode B or is there a difference?

Greetings,

Det

[jmbecker]

Mode A weights can be reported and assessed the same way as when using Mode B.

[JADI]

Hello,
I am currently dealing with SmartPLS Mode A/correlation weights. Specificallly, I am trying to justify the removal of some formative indicators.
In the SmartPLS handbook you refer to Cenfetelli & Basellier 2009 in this context. The authors state that “[t]he number of indicators has important implications for the statistical significance and, of course, the magnitude of the weights for those indicators". However, since I am using Mode A weights I am not sure if this citation is what I am looking for, i.e. an adequate justification for the removal of my indicators.
Can you please give me assistance regarding the mentioned statement/whether this is also valid when using Mode A weights? If not, it would be great if you could give me a hint how I could justify the removal of low/insignificant weights.

A further aspect I am interested in (even though I think that I know the answer) and the reason why I add my question to this thread concerns the interpretation of the Mode A formative weights.
I am dealing with formative indicators, thus it should be possible to interpret the weights from Mode A the way I can interpret the weights gained by using Mode B, right (positive weight means positive impact on the latent construct and finally on my endogenous construct (given a positive path coefficient between those constructs)?

I am looking forward to hearing from you.

Sincerely,

JADI

[jmbecker]

1)
Also for Mode A the general mechanism is true that adding indicators with non-zero weights will reduce (on average) the weights of the other indicators. This is particularly due to the standardization of the weights so that the .
However, in my opinion, you should always be careful with deleting indicators. If the indicators are theoretically relevant to the construct you should not just delete them. This could potentially bias results in other (future) studies that will (re-)analyze the same construct in other path models.

2)
Yes.