作者: R. Dennis Cook
DOI: 10.1080/01621459.1979.10481634
关键词:
摘要: Abstract Characteristics of observations which cause them to be influential in a least squares analysis are investigated and related residual variances, correlations, the convex hull observed values independent variables. It is shown how deleting an observation can substantially alter by changing partial F-tests, studentized residuals, variables, estimated parameter vector. Outliers discussed briefly, example presented.