Using PLS for my dissertation research

[emilio4]

PLS appears to be applicable to my disertation reserach data but the opinion of experienced PLS reserachers would be helpful.

I am attempting to test a path model which a large number of predictors but a small sample of 28 innovation team with 270 members. Potential predictors include: three sets of different diversity types of demographic, information and thinking style; team dynamics of learning, conflict, decision making,and team cohesion; and three leadership types. The dependent variale could be any of four including innovativeness and radical innovation from two sources, the team itself and stakeholders of the team.

I used SPSS and have created team level variables by aggregating members scores of factors idenitfied through Principle Factor Analysis The team level varialbes could become the inouts to PLS.

Thanks for any guidance.

Emilio De Lia
Rutgers University
USA

Is PLS apllicable for my research?

[Diogenes]

Hi Emilio,
If I understood correctly your model, you are working with 10 predictors and n=28.

SmartPLS could estimates the parameters, but the problem with so little sample is the statistical power. I done 3 tests with G*Power:

A priori
Medium effect size = 0.15 (R2 > or = 0.13 will be detected as significant)
Sig. = 0.05
Power = 0.8
Predictors = 10
N = 118 (computed by G*Power)


Post hoc
Medium effect size = 0.15 (R2 > or = 0.13 will be detected as significant)
Sig. = 0.05
Predictors = 10
N = 28
Power = 0.15 (computed by G*Power) – significant coefficients will not be detected.

Sensitivity
Sig. = 0.05
Power = 0.8
Predictors = 10
N = 28 (computed by G*Power)
Effect size = 0.93 (R2 > or = 0.48) – just huge effects will be detected as significant.

The problem will be: the non-significant coefficient means that there is no relation in the population or in fact there is but it is not detected (low power)?

I hope this help.

Bido

[Swandy Banta]

Hello,
Yes I need help with my theses analysis. I will be so pleased if i can get some help