Standardized weights

[antonioaguiar]

Hello to all,

I´m new at SmartPLS. I´ll appreciate if someone can help me with a basic doubt.

In order to assess, only, the measurement model (having 4 LVs; 3 Formative and 1 Relective) of a conceptual model, I was able to run a model using a dependent binary variable (DBV) that dependes on those 4 LVs. Even though I used Mean 0, Var 1 on Data Metric, I got weights greater than one.

I was expecting weights smaller than one since I used standardized data. How can I interpret this?


Best regads.
Antonio Aguiar

[christian.nitzl]

Hi,

This could be the case when multicollinearity is a problem.

Maybe the use of another weighting schema is a resolution e. g. the factor weighting. The factor weighting schema is more robust against mulicollinearity. You should also check your dataset for outliers.

Best regards,

Christian

[antonioaguiar]

[quote="christian.nitzl"]Hi,

This could be the case when multicollinearity is a problem.

Maybe the use of another weighting schema is a resolution e. g. the factor weighting. The factor weighting schema is more robust against mulicollinearity. You should also check your dataset for outliers.

Best regards,

Hi Christian,
Thank you very much for yhe reply. In fact I am using factor weighting schema. After your suggestion I used other schema but the results did not improve. Maybe I have to drop an indicator in order to reduce multicolinearity. What criteria should I use ?

Best Regards
Antonio

[christian.nitzl]

Hi Antonio,

I have often seen that authors report the correlation between the indicators. That could give you some hints where the problem could be. Another possibility is to check the variance inflation factor (VIF). For the calculation of VIF you have to use SPSS or another statistic program.

I hope that helps!

Christian

[antonioaguiar]

Thanks!
I will check the VIFs for each of the LV using SPSS. Hope to find the reasons why standardized weights are greater than one and eliminate the problem.

Kind Regards
Antonio