GOV 385L • Simultaneous Equations Methods
1:00 PM-2:30 PM
This course is about statistical models of more than one structural equation, accounting for more than one dependent variable. We shall concentrate on linear models of a standard econometric sort, although we shall also consider nonlinear models, models with discrete dependent variables, and models with measurement error as time permits. Computer exercises will provide practice at generating and interpreting concrete results, but the lectures and readings will dwell on more general questions of modeling, estimation, and inference: What sorts of models implyand should reflect hypotheses ofwhat sorts of effects? What variablesand equationsmust we include? What assumptions must we make, and what do they mean? How likely are the assumptions to be violated, and with what consequences? When is a model identified (roughly, estimable), and what can be done when it's not? What quantities should we be focusing on estimating? What estimators provide statistically desirable estimates? Where several different estimators might serve, what are their advantages and disadvantages? What do the estimates tell us, and how certainly? The answers enable us to understand the numbers the computer providesand to decide what to ask it to do in the first place. The lectures will treat these questions practically but abstractly, referring more to x's and y's than to substantive variables. There will be much mathematical notation and mathematically phrased argument and some proof and derivation. I hope to convey a good, relatively deep understanding of the hows and whys of constructing, estimating, and interpreting the estimates of these models.
Students taking this course should be familiar with at least the rudiments of matrix algebra and calculus, have a good knowledge of basic mathematical statistics (including familiarity with probability distributions, expected values, and their properties), have already taken Statistics II or equivalent, and be proficient with at least one suitable statistical software package, like SPSS, SAS, S+, or STATA.