The 'Nagelkerke R-squared' is suggested as a general measure of 'coefficient of determination' (Biometrika 78 (3): 691-692 SEP 1991) and is used by SPSS for logistic regression.
It is also possible to use predicted probabilities to derive an 'ROC' curve. Does anyone know if it there is a relation between the area under that curve and the Nagelkerke R-squared?
I thought I would ask before I set about trying to calculate it! A fairly simple search came up with nothing.
Thanks in advance
Phil McShane
Phil McShane
Nuffield dept of Surgery
John Radcliffe Hosp
Headington
Oxford
OX3 9DU
(+)1865 221312
fax (+)1865 768876
8 Manor Way
Kidlington
Oxon OX5 2BD
01865 378497
mob: 07791492029
|