Velilla Cerdan, SantiagoUniversidad Carlos III de Madrid. Departamento de Economía2008-08-192008-08-191990-102340-5031http://hdl.handle.net/10016/2812This paper presents an approximation for assessing the effect of deleting an observation in the eigenvalues of the correlation matrix of a multiple linear regression modelo Applications in connection with the detection of collinearityinfluential observations are explored.application/pdfengAtribución-NoComercial-SinDerivadas 3.0 EspañaCase deletionCollinearityEigenvaluesExtreme casesGateaux differentiabilityMultiple Linear RegressionPerturbation theoryworking paperEconomíaopen access