FormativeIndicators:NegativeWeight&/Loadings-Interpretation?

[HansDampf3000]

Hello,

i already asked the following question in another question but with a different title, so maybe it is better to open a new topic to get an answer:
I am evaluating a SEM with 3 formative constructs which are supposed to show me their effect on a single reflective construct with two indicators.
Regarding the reflective construct I checked for consistency reliability (Composite reliability) and convergent validity (Average Variance Extracted, Outer Loadings). The commonly in the context of reflective constructs conducted discriminant validity evaluation, is not necessary since only one reflective construct is given, right? Is an assessment of Q^2 necessary/possible when there is only 1 reflective goal construct?

Coming to my main question regarding the formative constructs:
Some indicators show a negative weight and/or a negative Loading (marked in yellow).
The items are coded correctly and the VIFs are all below 1.8, so there should be no problem with collinearity issues, right?
Since the formative indicator weights are all significant, can i interpret the results in a way as that they all contribute negative to the respective constructs?

It would be great if you could give me some advice how to ensure a correct interpretation of my results!

thank you in advance for your efforts!

Kind regards
Hans Dampf 3000

[HansDampf3000]

After some literature review I found the following in Cenfetelli 2009 (Interpretation of Formative Measurement in Information Systems Research MIS Quarterly Vol. 33 No. 4/December 2009)
"An indicator with a statistically significant negative weight but otherwise having a positive bivariate correlation with the formatively measured construct should be interpreted as an indicator having a negative effect when controlling for the effects of other indicators."
Is it relevant that the Loadings of the negative items are not significant?

But to be honest, I don't understand the last part of the sentence. Would you mind to tell me what "... effect when controlling for the effects of other indicators" means?
And is this interpretation also valid for the items with negative weights AND Loadings?

Kind regards
HansDampf

[cringle]

Hi

Do you expect all outer weights to be positive? If so:

(1) Check the correlation matrix of the indicators per formative measurement model. The correlations should all be positive. --> If not: data problem / or rescale

(2) If they are all positive, check the VIF of the formative indicators they should be (nicey) below 5. --> If not: collinearity problem (e.g., merge some indicators)

(3) If collinearity is not a problem, double click on the latent variable and select Mode A. Thereby, you use correlations weights to determine the formative construct.

Hopefully this hepls.

Kind regards
Christian

[ahg]

(3) If collinearity is not a problem, double click on the latent variable and select Mode A. Thereby, you use correlations weights to determine the formative construct.

Hi Christian,

Shouldn't that read "If collinearity is a problem"?

Best,

Ángel.

[cringle]

If collinearity is a problem you already found the problem. Then you use the usual strategies (delete an indicator, merge etc.). Here you could also try Mode A. But if you collinearity is not a problem you should give Mode A a try.

Kind regards
CR