Applied Econometrics I

Description

This is a course on Applied Regression Methods. Myriad regression models and techniques have become ubiquitous methods for eliciting useful information and identifying patterns and relationships in Data. Policy analysts of all stripes routinely use such methods in their work. In any regression model we have one dependent variable and some independent/predictor/explanatory vari ables. Our objective is to find out how much do the changes in the values of the independent/predictor/explanatory variables affect the change in the dependent variable. The Linear Regression Model (LRM) and the estimation technique of Ordinary Least Squar es (OLS) constitutes the basic building blocks of regression analysis. To be able to understand, learn , and use advanced and complex models of regression a good grasp of LRM and OLS is indispensable. In this semester we will cover LRM and OLS from an applied perspective. I start with the basics – Correlation and Scatter Plots. But we will build very fast on that. Next we cover bivariate regression (not that useful in practice, but quite useful as a sim ple pedagogic starting point) as an extension of the Scat ter Plots. Within bivariate regression we will emphasize the important differences in interpretation when the independent variable is an interval (or continuous) level variable and when it is a binary or dichotomous variable. After that we move over to multiple regression – which is what is really useful. We progress step by step using different types of independent variables – all interval level, all dichotomous , mixture of interval and dichotom ous, interaction terms (moderator analysis) involving different types of variables, and finally all types together. At this stage overwhelming emphasis will be on interpretation of the coefficient s and of the model – what story does our model tell? Later we will move on to hypothesis test and inference – to generalize to the population the results that we have got from our sample. Next we will cover some residual analysis and that will lead us to the pathological cases of breakdown of the assumptions of the Normal/Gaussian LRM. We will finish with some discussion of the causes, implications and remedies for some of these failures.