Seconnd ordered

[pantea]

Dear users,

I have got a big problem at the end of my thesis. I would be very thankful if anybody might help me...

I want to verify many hypotheses about correlations of latent variables. . all of my latent variales are formative . The output of PLS gives me information about the absolute correlations of the latent variables. but there are too many relations I want to reduce the relations by using second ordered construct because some constructs have form same meaning and form common concept for example constructs A,B,C,D form concept X ,I want to use second ordered to reduce my relations I used VPLS and it does not support boostrap for the second ordered model I did it manually by usuing the average data of indicators of each A,B,C,D and used a new construct which is a X for 3 other concepts I did the same and my model reduced to 5 constructs .and the t statistics are significant in some relations and my model is fitted

2 question rise up:

a)what I am using manually has statistical proof and does the same thing as second ordered model?


b) is it a better to use smart pls for this reason?

Thanks a lot for your help


Pantea

[viswadatta]

Second order constructs can be modelled using repeated indicator approach mentioned in this forum. Whether you should use it at all, depends on your theory building. The same set of statistics will be used for judging reliability and validity.

[Diogenes]

Hi all,
1) In fact, we should use repeated indicators in the 2nd order LV, just with reflective mode.
2) The use of averaged data is not wrong. And is a requested feature, see viewtopic.php?t=47
The only difference is that you are using the same weights for all indicators, regardless that some of them could be better than others.
3) It is a good idea compare the results of the complex model with the second.
Best regards.
Bido