[R] Step and AIC

[Steve Lianoglou]

Hi,

On Wed, Jan 27, 2010 at 12:21 PM,  <francesca.iordan at googlemail.com> wrote:
> Hello everybody,
> I would need some help from you.
> I am trying to fit a logistic model to some presence absence data of
> animals living on river islands. I have got 12 predictor variables and I am
> trying to use a stepwise forward method to fit the best logistic model to
> my data. I am using the function STEP (stats).
> I have a question for you. Can I use step function if my variables have a
> binomial distribution?
> Reading the explanations of the function, I have understood that step is
> more suitable for dealing with gaussian distributed variables.
> Is that right?
>
> I apologize in advance for this question, but I am just at the beginning of
> my long path to handle and know statistics and R.

This isn't really answering your question at all, but instead of doing
stepwise regression, could I recommend trying a regularized/penalized
regression model, instead?

The glmnet function in the glmnet package can fit the entire
regularization path for lasso/elasticnet -regularized logistic
regression model ... might be worth a try.

-steve

-- 
Steve Lianoglou
Graduate Student: Computational Systems Biology
 | Memorial Sloan-Kettering Cancer Center
 | Weill Medical College of Cornell University
Contact Info: http://cbio.mskcc.org/~lianos/contact