Core Methods: Foundations: Estimation and Identification
by Prof. Dr. Johannes Giesecke
This course will guide our PhD students through workhorse methods in cross-sectional, time-series and panel data analysis with specific attention to causal identification. This applied course begins with a review of OLS regression assumptions and consequences of violations with extensions into weighting, interactions and non-linearity. It then transitions into time-series and panel data analysis with attention to weak dependence, stationarity, cointegration and the estimation of short and long-term effects. Core panel designs, fixed and random effects, difference models and difference-in-difference designs also receive attention and application.
The PhD students also learn the advantages and challenges of multilevel analysis and random coefficient models. Selection effects and endogeneity issues receive extended attention and participants are introduced to common solutions in the form of Heckman models and instrumental variable regression. The final weeks of the course turn to maximum likelihood estimation and categorical data designs. This course places a strong emphasis on actual application. All classes take place in the computer lab and divide time between theory and application. The PhD students are assigned a problem set at the end of each class covering that day’s materials and the beginning of the following class is used to review the answers. As this is an applied class, most of these assignments involve the proper analysis of practice datasets using R. In addition, participants are also offered a weekly tutorial in R programming techniques necessary to complete the given week’s problem sets.