Re: [R] help to setting a multiple (linear) regression model with a 5% significance level (threshold) for the inclusion of the model variables.

[Prof Brian Ripley]

On 22/11/2011 16:30, gianni lavaredo wrote:
Dear Researchers,
someone know the right syntax to chose a 5% significance level (threshold)
for the inclusion of the model variables in a  multiple (linear) regression
in backward way?

I set the formula in this way, but I don't know to choose the 5%
significance?

lmodelV<-
step(lm(formula=MyFormula[[1]],data=LR.train),direction="backward")

That is not what step() does.  Please do read the help page!

Description:

     Select a formula-based model by AIC.

Note, 'by AIC'.

Now, as Frank Harrell says here often, stepwise methods are inadmissible 
for prediction, and in addition stepwise F-test methods are always 
invalid as they ignore multiple testing.  So you won't find such methods 
prominently available in R (and all the stepwise methods I am aware of 
in R packages use AIC or similar).

The author of step()

thanks in advance gianni

        [[alternative HTML version deleted]]

Please do not send HTML when expressly asked not to.



______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.


--
Brian D. Ripley,                  rip...@stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595

______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.