Hi,
I am using rms R-package for a Binary Logistic Regression.
Is it possible to get Logit Rank Plot explained by J.Copas in rms?
Definition of Logit Rank Plot:
Logit Rank Plot has been introduced by j.Copas in his paper “The effectiveness of risk scores: the logit rank plot”. It is an alternative way to assess the discimination power of a a fitted logistic regession model.
As described in the paper the Logit Rank Plot is the plot of
logit(p)=log{p/(1-p)}
agaist
logit(F(s)=log(F(s)/(1-Fs)
where logit(F(s)=r is the proportional rank of s in a sample or population; where s is the risk score for the event E= “default or not default”
The logit rank plot provides a curve; the steeper is this curve, the greater the degree of discrimination between high and low risk cases.
"The logit rank plot gives a complete description of the joint distribution of s and the binary
outcome E/E, up to an arbitrary monotonic transformation of s. The ROC curve, by con-
trast, only describes the conditional distributions of s given E and E, again up to a monotonic
transformation of s. The ROC curve can be reconstructed from the logit rank plot but not
vice versa."