Hello,
I have a first-order but multidimensional (3 dimensions and each dimension has 3 items, totally 9 items) reflective construct. Before that I tried it as a second-order construct, but one of Smartpls developers in my country told that because of the collinearity issues, you should make it as a first-order.
My model is now reflective to reflective.
After doing consistent pls and bootstrapping, sample mean and stdev results all are n/a. T statistics are all 0.
Then, if I try non-consistent bootstrapping, I get a significant hypothesis result among all three hypotheses on this reflective-reflective model.
All the constructs are reflective. I know that I should use consistent algorithms. In my case, can I use non-consistent version to write in my thesis or should I change the first-order and multidimensional construct to formative one?
If I do the analysis making it formative one, I get the same result (a significant hypothesis result among all three hypotheses).In case of formative analysis, do I need to make redundancy analysis for the first-order and multidimensional formative construct?
Best regards.
Tugberk, MBA student