interpretation of dummy variables (exogenous) with "lavaan"

[Gary Dong]

Dear all,

I wonder how I should interpret the standardized coefficients of dummy variables in SEM. I have an exogenous categorical variable with three levels: low, medium, and high. According to the online tutorial for "lavaan" (http://lavaan.ugent.be/tutorial/cat.html), I simply re-code it into three dummy (0/1) variables and include two of them in the model. Now I wonder how I should interpret the standardized coefficients of the two dummy variables.

For example, when I use "medium" as the reference, the standardized coefficient for "low" is 0.40 and the standardized coefficient for "high" is -0.20. 

How should I interpret them? Is it correct to interpret them as I do for continuous variables (one-unit std. deviation in X is associated with one-unit std. deviation change in Y)?

Your advice is greatly appreciated.

Best

Gary

[Mikko Rönkkö]

Hi,

Standardized estimates of dummy variables are difficult to interpret, so you should probably use unstandardized estimates instead. The standardized estimates dependent on the group frequencies, and that is probably not something that you want.

Once you have unstandardised the estimates, the dummy variables are interpreted as mean differences between the two other groups and the reference four. 

Mikko