R help - Mar 2009 - - help - LARS & LASSO based on glm

Hello

I'm trying to use LARS on a glm:

    log.1 = glm(Train.Data$TL~(.),Train.Data,family = binomial)

    LARS.1 = lars(model.matrix(log.1),Train.Data$TL,type =
"lar",max.steps ncol(model.matrix(log.1))-1)

    self.fit
predict.lars(LARS.1,model.matrix(log.1),type="fit",mode="step")$fit

    summary(self.fit)

just even with a self fit, I get values not exactly between 0 and 1 (max is
always 1, but the minimal is around -.007 for most of the 58 steps of the
LARS)

how can I fix it so that the LARS predicts values between 0 and 1 ?

thanks

eitan

	[[alternative HTML version deleted]]

> how can I fix it so that the LARS predicts values between 0 and 1 ?
I don't know, but you can use glmnet to accomplish the same thing.

-- 

Max