香蕉视频

Davidson Distinguished Lecture in Econometrics with Prof. James MacKinnon

"Using Large Samples in Econometrics"

(Queen's University)
October 13, 2023, 3:30 to 5:30 PM
Faculty Club Ballroom

Abstract:
The samples that economists employ have become larger over time, sometimes a lot larger. This seems good, because large samples generally contain more information than small ones. However, inference in large samples can actually be harder than in small ones. Hypothesis tests and confidence intervals are often based on the assumption that every observation is independent of every other observation. But violations of this assumption are far more consequential for large samples than for small ones. If the observations naturally fall into a modest number of clusters and are correlated within each cluster, then the information content of a large sample may be only slightly greater than that of a much smaller sample. Moreover, failing to account for clustering can lead to much bigger mistakes when a sample is large than when it is small. These points are illustrated graphically using both simulated and actual data.

Registration:
This event is free and is open to anyone interested in the topic. However, due to limited seating, registration is required.聽The deadline to register is聽October 9 at 5 pm.