Outlier Detection and Removal for Econometrics Models
Anomaly detection is a major machine learning technique aiming to identify the deviant samples from the general data distribution. It has many key business applications in various fields, including intrusion detection, mechanical fault detection, credit card default prediction, and terrorism event detection. Other than simple outlier detection methods, e.g., removing the samples which are more than two standard deviations from the sample mean, many more advanced and complex approaches are proposed in the past several decades. However, their usages in traditional econometrics field is still rare, although many models are sensitive to outliers. It is acknowledged that the presence of outliers may skew the estimation result heavily. In this study, we will empirically examine the effect of latest outlier detection algorithms on classical econometrics models.
This is an ongoing project. Use with Caution :)