This paper discusses a number of areas of inference where dissatisfaction by applied workers with the prescriptions of econometric high theory is strong and where a likelihood approach diverges strongly from the mainstream approach in its practical prescriptions. Two of the applied areas are related and have in common that they involve nonstationarity: macroeconomic time-series modeling, and analysis of panel data in the presence of potential nonstationarity. The third area is nonparametric kernel regression methods. The conclusion is that in these areas a likelihood perspective leads to more useful, honest and objective reporting of results and characterization of uncertainty. It also leads to insights not as easily available from the usual perspective on inference.
All Science Journal Classification (ASJC) codes
- Economics and Econometrics
- Kernel regression
- Panel data