problem with multiple exogenous variables vs. one single exo

[Sam]

Hi all,

my model consists of a latent second order endogenous variable and several latent exogenous variables (Ex1 to Ex5), that are only connected to the endogenous variable. If I only implement one exogenous variable in the model, the path coefficients look like this for example (results from several seperate models):
Ex1: .334, p<.05
Ex2: .128, p<.05
and so on.

If I include all of the exogenous variables in the model and each variable only influences the endogenous construct, path coefficients look like this:
Ex1: .011, n.s.
Ex2: -.127, p<.05
and so on.

That means, some strong influences are not significant any more and others change their direction from positive to negative.

Has anybody an idea what the problem is? All constructs have three indicators and are reflective. All quality criteria named by Hair et al. 2014 are met well.

Thanks for any hints in advance!

Sam

[cringle]

Hi, that's not a problem but a typical result. The more exogenous variables you include, the lower the path coefficients (if you only have one, it's the strongest). If, in addition, collinearity becomes an issue, signs may also change.

Best
CR