Hello!
Is multicollinearity possible in SEM research? I have six latent variables in my hypothesized research model, and four of these variables hypothesized to have direct effects on two other variables.
FYI:
A, B, C, D are exogenous, and E, F are endogenous variables. A, B, C, D are hypothesized to have a direct effect on E and F, and E is hypothesized to have a direct effect on F. So it is a saturated model.
The two endogenous variables (E and F) appear to have a high correlation according to the initial measurement model results: .876. My confusion is that one exogenous variable has a positive correlation with an endogenous variable (A with F), however, when I conduct the structural model assessment, it has a negative effect on the endogenous variable. My suspicion was that the two endogenous variables (E and F) were causing multicollinearity in the model, so I removed the variable E from the model just to see if the sign was going to change to positive. However, the sign remained negative after I removed the variable E from the model and conducted the assessment again. Is this normal in SEM research - to have a positive correlation but a negative direct effect? Does it mean that there is no multicollinearity in the model because the sign remained the same after I removed the variable E?
Dear Yalcin
To put it simple, YES, multicollinearity is possible in SEM.
When you suspect high correlation between your measured variables, you might want to include the residual correlations when specifying the models. A good starting point is to make your model saturated by including all the important linkages or processes in your model. This means that you incorporate all the three components of the model (1. Definition of latent variable, 2. Specification of the regressions and 3. Specification of the residual correlation).
Your definition of latent variable is however confusing, as latent variable (some say composite variable) is usually composed of two or more measured variables. So you define yourself your latent variables from combination of other variables.
For available resources, kindly explore the lavaan package by Yves Rosseel and resources by Jim Grace.
Article Tree species diversity promotes aboveground carbon storage t...
Multicollinearity is a data problem that can adversely impact regression interpretation by limiting the size of the R‐squared and confounding the contribution of independent variables. For this reason, two measures, tolerance and VIF, are used to assess the degree of collinearity among independent variables.
- Badges
- Science topic
- Similar topics
- Physical Sciences
- Materials Science
- Materials
- Inorganic Materials
- Ceramics