Hello everyone,
can someone help me with the following problem:
I have run a model with three predictor variables, one moderator variable, one dependent variable and three moderating effects (using standardized indicators).
For one of the moderating variables I obtained unsatisfactory composite reliability (0.0038), AVE (0.1856), and factor loadings (between -0.3 and 0.57).
How would you deal with that? Would you delete the whole moderating effect or first try to eliminate some of the manifest variables (which are obtained as products from the manifest variables of the predictor and the moderating variable), as these indicators are by definition parallel measures of the underlying latent interaction variable?
Thanks for your advice!
Unsatisfactory reliability outcomes of moderating effects
if i understand your problem correctly, your have 5 variables, as follow:
P1, P2, P3, M1, and D1. There're 4 main effects from P1, P2, P3, and M1 going into D1; and 3 moderating effects P1*M1, P2*M1, and P3*M1.
When you say "For one of the moderating variables I obtained unsatisfactory composite reliability (0.0038), AVE (0.1856), and factor loadings (between -0.3 and 0.57)"... do you mean the reliability for M1, or one of the P*M1?
As far as I know, you need to worry about the reliability of the variables, i.e. P1, P2, P3, M1, and D1; but not the interaction of variables.
P1, P2, P3, M1, and D1. There're 4 main effects from P1, P2, P3, and M1 going into D1; and 3 moderating effects P1*M1, P2*M1, and P3*M1.
When you say "For one of the moderating variables I obtained unsatisfactory composite reliability (0.0038), AVE (0.1856), and factor loadings (between -0.3 and 0.57)"... do you mean the reliability for M1, or one of the P*M1?
As far as I know, you need to worry about the reliability of the variables, i.e. P1, P2, P3, M1, and D1; but not the interaction of variables.
- ghozali
- PLS Expert User
- Posts: 39
- Joined: Sat Oct 15, 2005 1:18 am
- Real name and title: Prof. Imam Ghozali, Ph.D
- Location: Indonesia
Hi,
If you have moderating variable it means that your model is not linear because you have high order model with interaction. In term of regression the independent variable become un-intepretable so you don't have to worry with independent variable. Only the interaction variables can be interpreted.
If you have moderating variable it means that your model is not linear because you have high order model with interaction. In term of regression the independent variable become un-intepretable so you don't have to worry with independent variable. Only the interaction variables can be interpreted.
Faculty of Economics, Diponegoro University
Jl. Erlangga Tengah 17 Semarang, Indonesia
ghozali_imam@yahoo.com
Jl. Erlangga Tengah 17 Semarang, Indonesia
ghozali_imam@yahoo.com