I’m doing Structural Equation Modeling using Lisrel. I basically want to do a path model involving both observed and latent variables at the same time. The model that I want to test is a mediation model, meaning that I want to test if X => M (Mediation variable, obs. variable) => Y (Latent variable).

I simplified the model (see below), it contains three observed variables X1 – X3, two observed variables that should be the mediators, M1 and M2, and three observed variables Y1 - Y3 representing the outcome and together form one latent variable (WBEING). The syntax is as follows:

Model WBEING

Observed variables:

X1 X2 X3 M1 M2 Y1 Y2 Y3

Correlation matrix:

….

Sample size: 311

Latent variables: WBEING

Relationships:

Y1 = 1*WBEING

Y2 = WBEING

Y3 = WBEING

M1 = X1 X2 X3

M2 = X1 X2 X3

WBEING = M1 M2

Path diagram

End of problem

I get the following Warning message, that I cannot figure out, because it does not seem to make any sense, as the indicated path is not defined in the syntax:

W_A_R_N_I_N_G : The path from X1 to Y1 could not be interpreted.

I do not get this message when I work with latent variables, ie. create a latent mediating variable M (consisting of M1 and M2), so I guess the mistake could be that you cannot make such a model using observed variables. I cannot find any information in the Lisrel User’s guide nor in the web.

Does anyone have an idea? Thanks a lot in advance for every comment! :wave: