A multi-row deletion diagnostic for influential observations in small-sample regressions

作者：Kaffine Daniel T; Davis Graham A

来源：Computational Statistics & Data Analysis, 2017, 108: 133-145.

DOI：10.1016/j.csda.2016.10.007

• Regression diagnostics
• Multi-row deletion
• Inference
• Influence points
• Outliers
• Growth regressions
• Institutions
• Resource curse

The inference from ordinary least-squares regressions is often sensitive to the presence of one or more influential observations. A multi-row deletion method is presented as a simple diagnostic for influential observations in small-sample data sets. Multi-row deletion is shown to be complementary to two related diagnostic tests, DFBETAS and robust regression. As an illustration, the technique is applied both to simulated data and to a real data set from an influential study examining the role of institutions for economic growth in resource-rich economies. Multi-row deletion reveals that the key economic insight that institutions matter is sensitive to small variations in sample, indicating additional analysis may be warranted.

• 出版日期2017-4