Hello there,
Hope someone can help me with the following question.
I have a reflective-formative model where several reflective latent constructs, such as B, C, and D, affect a formative construct A, which consists of A1, A2, and A3 which are latent reflective variables. Given the fact that this formative construct A is an endogenous variable in the model, i can't use the usual way to model A.
I heard about some two-stage approach. So I ran the model by setting up A (with all the indicators of A1, A2, and A3 used as A's own indicator in a reflective manner) as being formed by A1, A2, and A3, along with all other exogenous variables that cast impacts on A. Based on this model, I calculated the latent scores of A1 (noted as A1^), A2 (A2^), and A3 (A3^). Next, I removed latent reflective constructs A1, A2, and A3 from the model. I also removed all the indicators of A. Finally, I added A1^, A2^, and A3^ as the only 3 indicators of A in a formative manner.
Now I ran the model again but to see poor weights and negative weights of A1^, A2^, and A3^ on A. I wonder whether I did anything wrong?
3 weeks ago, I did the same approach to analyze a research model of a completely different research project with completely different data. I also observed very poor weights of the transformed indicators. I am very upset.
Please help!!! Thanks so much.
Two-Stage Approach on Reflective-Formative Model
-
Hengkov
- PLS Super-Expert
- Posts: 1599
- Joined: Sun Apr 24, 2011 10:13 am
- Real name and title: Hengky Latan
- Location: AMQ, Indonesia
- Contact:
Hi Pae,
Your model have effect interaction or not => Two-Stage Approach?
You construct A is second order Type II (A1, A2, A3 to A)?
I see you explain above, you model have 3 exogenous variabel reflective (B,C,D) and 1 endogenous variabel formative (A => with 3 construct A1,A2,A3 to A => mode B), please correct.
For estimate this model run algorithm PLS, evaluation outer model construct (first-order), for construct A (if second order used repeated indicators), next run bootstrap procedure.
Regards,
Hengky
Your model have effect interaction or not => Two-Stage Approach?
You construct A is second order Type II (A1, A2, A3 to A)?
I see you explain above, you model have 3 exogenous variabel reflective (B,C,D) and 1 endogenous variabel formative (A => with 3 construct A1,A2,A3 to A => mode B), please correct.
For estimate this model run algorithm PLS, evaluation outer model construct (first-order), for construct A (if second order used repeated indicators), next run bootstrap procedure.
Regards,
Hengky
re:
Hello Hengkov,
Thank you for your response.
Quote from Ringle et al., A Critical Look at the Use of PLS-SEM in MIS Quarterly, MIS Quarterly, 36 (1), pp. iii-xiv, 2012 :
"In this kind of situation (a reflective-formative model; a formative construct consists of several reflective constructs; this formative construct is an endogenous variable and other exogenous variables predict this formative construct); , a mixture of the repeated indicators approach and the use of latent variable scores in a two-stage approach—which
is similar to the two-stage approach in moderator analyses in PLS-SEM (Henseler and Chin 2010)—is appropriate. In the first stage, one uses
the repeated indicators approach to obtain the latent variable scores for the lower order components which then, in the second stage, serve as
manifest variables in the measurement model of the higher order component (Figure B2). Thereby, the higher order component is embedded
in the nomological net in a way that allows other latent variables as predecessors to explain some of its variance, which may result in significant
path relationships."
This is the two-stage approach that i mentioned. It is not about interaction testing. My apologies for the confusion.
And, A is a second order formative constructs with A1, A2, and A3 as the underlying constructs. A1, A2, and A3 are all reflective constructs as each has several indicators.
I don't fully understand your suggestion. Do you mind explaining it with more details? I mentioned in my original post that I did the following analyses. I felt that what I did was somewhat similar to your suggestion.
"So I ran the model by setting up A (with all the indicators of A1, A2, and A3 used as A's own indicators in a reflective manner -- which is the repeated indicator approach that i understand) as being formed by A1, A2, and A3, along with all other exogenous variables that cast impacts on A. Based on this model, I calculated the latent scores of A1 (noted as A1^), A2 (A2^), and A3 (A3^). Next, I removed latent reflective constructs A1, A2, and A3 from the model. I also removed all the indicators of A. Finally, I added A1^, A2^, and A3^ as the only 3 indicators of A in a formative manner. "
Again, I am in a great need of help. I can't move on my project unless i resolve this issue. Many thanks to those who can help!!
Thank you for your response.
Quote from Ringle et al., A Critical Look at the Use of PLS-SEM in MIS Quarterly, MIS Quarterly, 36 (1), pp. iii-xiv, 2012 :
"In this kind of situation (a reflective-formative model; a formative construct consists of several reflective constructs; this formative construct is an endogenous variable and other exogenous variables predict this formative construct); , a mixture of the repeated indicators approach and the use of latent variable scores in a two-stage approach—which
is similar to the two-stage approach in moderator analyses in PLS-SEM (Henseler and Chin 2010)—is appropriate. In the first stage, one uses
the repeated indicators approach to obtain the latent variable scores for the lower order components which then, in the second stage, serve as
manifest variables in the measurement model of the higher order component (Figure B2). Thereby, the higher order component is embedded
in the nomological net in a way that allows other latent variables as predecessors to explain some of its variance, which may result in significant
path relationships."
This is the two-stage approach that i mentioned. It is not about interaction testing. My apologies for the confusion.
And, A is a second order formative constructs with A1, A2, and A3 as the underlying constructs. A1, A2, and A3 are all reflective constructs as each has several indicators.
I don't fully understand your suggestion. Do you mind explaining it with more details? I mentioned in my original post that I did the following analyses. I felt that what I did was somewhat similar to your suggestion.
"So I ran the model by setting up A (with all the indicators of A1, A2, and A3 used as A's own indicators in a reflective manner -- which is the repeated indicator approach that i understand) as being formed by A1, A2, and A3, along with all other exogenous variables that cast impacts on A. Based on this model, I calculated the latent scores of A1 (noted as A1^), A2 (A2^), and A3 (A3^). Next, I removed latent reflective constructs A1, A2, and A3 from the model. I also removed all the indicators of A. Finally, I added A1^, A2^, and A3^ as the only 3 indicators of A in a formative manner. "
Again, I am in a great need of help. I can't move on my project unless i resolve this issue. Many thanks to those who can help!!
Pae,
From what I understand you have an endogenous Type II construct (which you call A) and some other exogenous constructs. Construct A is second order, formed by three first order factors that have reflective indicators. I think what you need to do is the following. First, use the repeated indicator approach which means you take all the indicators of the three sub-constructs of A (A1, A2, and A3) and link them to your 2nd order construct A in a reflective way. So if each of the three sub-constructs had 3 indicators, you end up with 9 indicators connected to your second-order construct A. You then run the whole model and obtain the factor scores. Then in the 2nd step, you link the 2nd order construct A directly to the three factor scores of A1, A2, and A3 in a formative way. You also use the other factor scores as indicators of all other constructs in your model. Run the model again to obtain the path coefficients, then bootstrap for significance. You need to also do the standard reliability and validity tests of course. Hope this helps.
Shamel
From what I understand you have an endogenous Type II construct (which you call A) and some other exogenous constructs. Construct A is second order, formed by three first order factors that have reflective indicators. I think what you need to do is the following. First, use the repeated indicator approach which means you take all the indicators of the three sub-constructs of A (A1, A2, and A3) and link them to your 2nd order construct A in a reflective way. So if each of the three sub-constructs had 3 indicators, you end up with 9 indicators connected to your second-order construct A. You then run the whole model and obtain the factor scores. Then in the 2nd step, you link the 2nd order construct A directly to the three factor scores of A1, A2, and A3 in a formative way. You also use the other factor scores as indicators of all other constructs in your model. Run the model again to obtain the path coefficients, then bootstrap for significance. You need to also do the standard reliability and validity tests of course. Hope this helps.
Shamel
-
Hengkov
- PLS Super-Expert
- Posts: 1599
- Joined: Sun Apr 24, 2011 10:13 am
- Real name and title: Hengky Latan
- Location: AMQ, Indonesia
- Contact:
Hi Pae,
Your case is similar with Soliman Moataz some days ago.
You have endogenous second-order Type II (Reflective-Formative: A1,A2,A3 to A).
Ringle et al. (2012) => Appendix B explain Two-Stage Approach for Type I and Type IV (Shamel explain above for Type II). ;-)
You result weight is negative (before you delete some indicators A1,A2,A3 and A => outer loading < 0.5?).
Weight negative is same positive.
(T -1.96).
Regards,
Hengky
Your case is similar with Soliman Moataz some days ago.
You have endogenous second-order Type II (Reflective-Formative: A1,A2,A3 to A).
Ringle et al. (2012) => Appendix B explain Two-Stage Approach for Type I and Type IV (Shamel explain above for Type II). ;-)
You result weight is negative (before you delete some indicators A1,A2,A3 and A => outer loading < 0.5?).
Weight negative is same positive.
(T -1.96).
Regards,
Hengky