You can deal with multicollinearity in SEM by creating relationships between them (e.g. correlation or causation) or using a latent variable to eliminate spurious relationship.

You can deal with multicollinearity in SEM by creating relationships between them (e.g. correlation or causation) or using a latent variable to eliminate spurious relationship.

Another way of dealing with collinearity is by combining the variables. For example, if you have 5 variables and 2 of then are correlated in such a way that they are measuring the same thing then combine those 2. It will leave you with three variables. This works well for latent variables.

Dear folks,

I would like to note that colinearity increases the standard errors of parameter estimates but leaves the consistency and unbiasedness of these parameter estimates untouched.

Hence, what ever you to consider if the method may reduce colinearity but at the same time biasing and destroying the main theoretical models.

Approaches like PCA regression, PLS regression or combining variables lead to a changed causal model that isn't the theoretical model any more that the researcher once intended. In my research I firstly strive for a causal model that reflects my thinking of the field and I try to defend that model as long as i can against any data manipulation strategies or statistical approaches that change the model.

Just my 2c.

Best,

Holger

Multicollinearity refers to a situation in which or more predictor variables in

a multiple regression Model are highly correlated.

Ridge regression is one of the most important method which can be used to solve this problem because OLS estimators will not has BLUE properties.

you can use this URL to explain that clearly.

http://www.m-hikari.com/ijcms-2011/9-12-2011/rashwanIJCMS9-12-2011.pdf

the above URL shows that these important notes:

Multicollinearity has several effects, these are described as follow :

- High variance of coefficients may reduce the precision of estimation.

- Multicollinearity can result in coefficients appearing to have the wrong sign.

- Estimates of coefficients may be sensitive to particular sets of sample data.

- Some variables may be dropped from the model although, they are important

in the population.

- The coefficients are sensitive of to the presence of small number inaccurate

data values (more details in Judge 1988, Gujarat; 1995).

Collinearity is a consequence of the (very common) violation of one of the most fundamental assumptions underlying regression--that the X or predictor values are chosen, not merely sampled or observed like the Y or dependent values. If you could overcome the design flaw that left you with Collinearity / sampled X values in the first place, that would address the problem much better than an ex post adjustment. The after-the-fact fix is trying to answer an impossible question--what would your data be like if the research design was different than what it actually was?

Did u find a way to solve it????

You can also try using penalized maximum likelihood estimation approach.

The sample size can be increased if it has to do with primary data collection. Also, the number of independent variables can be increased.