Auxiliary Variables in a Longitudinal Model

When you look at the path diagrams of what we just described, actually your heart rate should go up a little bit because it looks like you're violating Gauss Markov. The term saturated as in saturated correlates implies that it is related to everything but in a way that doesn't otherwise disturb what you have. So one of the key assumptions we invoke in a general linear model is your predictors are not correlated with the disturbance, they're independent. That's the whole endogeneity problem that Conometrics deals with.