Showing posts from January, 2018

The Covar Matrix obsession (JM Martin)

No obsession is healthy, including statistical ones. But that doesn't mean that the Variance-Covariance Matrix should be left without deep analysis. However, more often than expected, Covar analysis takes many weeks of deep theoretical demonstrations, more weeks of hypothetical cases and few more of sample calculations. All this before actual and practical issues during the estimation of real models. Again, it's by all means important to assert this sort of issues that could seriously affect error term properties, as well as coefficients'. Thus, the theoretical level of which this milestone has turned into in the classroom, is far beyond what is academically sound and healthy for economics students. Therefore, econometric teachers need to show visually, by graphs that is, what are the different effects of non-optimal covar matrices, over forecasts and simulations. Sure, numbers and graphs can be hidden or customized in papers, but at least we are adding more prac