Is Your Structure a Mediation Model?

If your goal is measurement, don't screw with the measurement, got it? But if your goal is to understand the relations among the structural aspects of a model, so you might have a collection of latent variables that might be hypothesized to influence some others. The parceling in theory has a bearing on both of those. So now if we think about what happens in terms of parceling, when you take your variables and smush them together, you are changing the loadings. If you change the loadings, that may well change how stable a given factor is.