Help please
In my model I have employee motivation as a formative which has 2 dimensions (Intrinsic Motivation and Extrinsic Motivation) but based on my data, Intrinsic motivation is not significant.I am using SmartPLS, after bootstrapping(5000) still the the intrinsic motivation weight is below 1.96. But the loading factors are good. Please check the attachment.
Please suggest me what I should do.
Thank you
If you run AMOS on your model and check all the necessry indices, this will give you better idea about the structure of the model, and if there is a need to modify or delete some constructs.
Regards
Be sure that the non significance of the construct is not as a result of use of wrong model. if you are sure of that, then you can make use of only the ones that sre significant. that is the essence of variable selection.
If you run AMOS on your model and check all the necessry indices, this will give you better idea about the structure of the model, and if there is a need to modify or delete some constructs.
Regards
Dear
The structure of the model (formative or reflective) is based on previous theoretical justification. Therefore, if you have the theoretical justification about employee motivation as formative construct, then you need to check the necessary indices (e.g., CFI, RSMEA, NFI etc).
No you should delete the dimension just because the construct is not significant.
But you said you tested the dimension. There is no such this as testing the dimension because the hypothesis testing is about the causal effect of one construct (IV) on another construct (DV) while dimensions are the measure for the respective construct.
Dear
I think I found the exact answer to the above question as mentioned below:
A final note of caution: formative indicators should never be discarded simply on the basis of statistical outcomes. Such actions may substantially change the content of the formative index (Henseler et al., 2009; Jarvis et al., 2003). Thus, the researcher should keep both significant and insignificant formative indicators in the measurement model as long as this is conceptually justified. Usually, PLS structural model estimates hardly alter after performing an elimination of insignificant or highly collinear formative indicators, providing further support for the decision to retain such indicators in the PLS path model. In the case of multicollinearity, indicator weight estimates can be distorted. This fact requires researchers to be particularly cautious when interpreting indicator weights as a sign of indicator importance (Henseler et al., 2009).
Thank you
- Badges
- Science topic