Formative measurement model assessment
Posted: Thu Jul 04, 2013 11:15 am
Dear all,
Based on Hair et al (2014), A Primer On PLS-SEM, Chapter 5. There are 3 steps to do the validity assessment: 1) Assess convergent validity based on redundancy analysis, 2) Assess collinearity issues/VIF, 3) Assess significance and relevance via bootstrapp.
[b]My question is dedicated to step no 1 - convergent validity based on redundancy analysis.[/b] In Hair et al (2014), redundancy analysis is carried out by creating new model for the formative constructs and create one reflective construct with a global single-item measures which summarizes the essence of the formative indicators (Sarstedt, Wilczynski, Melewar, 2013).
Therefore, my concern is - what if we don't have that summarized single item in the survey? Again based on Hair et al (2014), we should include this global single item in the survey if we are to include formative indicators for the constructs. So,
i. How are we going to do the redundancy analysis?
ii. Can we just check the weight of the indicators and their t-values and delete the items with insignificant values?
iii. then after no (ii) we proceed with step 2 which is to assess the collinearity issues/VIF?
The main problem here is I don't have that global single-item measures and would not be able to get the values of convergent validity's path coefficient. :-(
Kindly advise and thank you in advanced.
Raha
Based on Hair et al (2014), A Primer On PLS-SEM, Chapter 5. There are 3 steps to do the validity assessment: 1) Assess convergent validity based on redundancy analysis, 2) Assess collinearity issues/VIF, 3) Assess significance and relevance via bootstrapp.
[b]My question is dedicated to step no 1 - convergent validity based on redundancy analysis.[/b] In Hair et al (2014), redundancy analysis is carried out by creating new model for the formative constructs and create one reflective construct with a global single-item measures which summarizes the essence of the formative indicators (Sarstedt, Wilczynski, Melewar, 2013).
Therefore, my concern is - what if we don't have that summarized single item in the survey? Again based on Hair et al (2014), we should include this global single item in the survey if we are to include formative indicators for the constructs. So,
i. How are we going to do the redundancy analysis?
ii. Can we just check the weight of the indicators and their t-values and delete the items with insignificant values?
iii. then after no (ii) we proceed with step 2 which is to assess the collinearity issues/VIF?
The main problem here is I don't have that global single-item measures and would not be able to get the values of convergent validity's path coefficient. :-(
Kindly advise and thank you in advanced.
Raha