Displaying 1 result from an estimated 1 matches for "ols0".
Did you mean:
ols
2010 Jun 29
1
Model validation and penalization with rms package
...the case that the optimum penalization may still
give a model which is substantially overfitted?
I calculated corrected R^2, optimism in R^2, and corrected slope for
various penalties for a simple example:
x1 <- rnorm(45)
x2 <- rnorm(45)
x3 <- rnorm(45)
y <- x1 + 2*x2 + rnorm(45,0,3)
ols0 <- ols(y ~ x1 + x2 + x3, x=TRUE, y=TRUE)
corrected.Rsq <- rep(0,60)
optimism.Rsq <- rep(0,60)
corrected.slope <- rep(0,60)
for (pen in 1:60) {
olspen <- ols(y ~ x1 + x2 + x3, penalty=pen, x=TRUE, y=TRUE)
val <- validate(olspen, B=200)
corrected.Rsq[pen] <- val["R-square&...