#standard errors

“In general we conclude that using ML methods for the analysis of multilevel data with non-normally distributed group level residual errors only causes problems when one is interested in the significance or in the confidence intervals of the variance terms at the second level.”

This is a [good post](http://andrewgelman.com/2011/10/how-do-you-interpret-standard-errors-from-a-regression-fit-to-the-entire-population/) by Andrew Gelman to keep in the back of my mind when when doing research with a sample that is the population (i.e. *all* Federal Reserve Chairman speeches).