too many var in lm

Thanks Eik for your reply.

Sure I know the classification method. However, at this stage, I'm working on feature selection. So lmfit and eBays in Limma are more preferable than anova in stats? 


Best wishes,

Haleh



----- Original Message -----
From: Eik Vettorazzi <E.Vettorazzi at uke.uni-hamburg.de>
To: carol white <wht_crl at yahoo.com>
Cc: 
Sent: Thursday, August 18, 2011 10:56 AM
Subject: Re: [R] too many var in lm

Hi Carol,
methods for classifying observations are legion, starting from logistic
regression, discriminant analysis, CART, (hierarchical) cluster analysis...

When it comes to analysing gene expressions, www.bioconductor.org might
be the place to visit, especially the limma-package might be promising.

Regards,
Eik


Am 18.08.2011 10:30, schrieb carol white:

Thanks Eik for your reply.

I have seen that one way to select variables discriminating 2 categories of patients is anova. I saw that the anova function should be applied to an object like the one obtained from lm. so that's why I wanted to apply anova(lm(y~.)). Would you have any suggestions, comments?

Regards,

Carol



----- Original Message -----
From: Eik Vettorazzi <E.Vettorazzi at uke.uni-hamburg.de>
To: carol white <wht_crl at yahoo.com>
Cc: "r-help at stat.math.ethz.ch" <r-help at stat.math.ethz.ch>
Sent: Wednesday, August 17, 2011 3:39 PM
Subject: Re: [R] too many var in lm

Hi Carol,
it might be another question if it is sensible to use 2100 regression
parameters, but you can use . to regress one response against all other
variables in a data frame as in:

lm(formula = mpg ~ ., data = mtcars)

and you can even exclude specific variables using "-"
lm(formula = mpg ~ . - wt, data = mtcars)

cheers.

Am 17.08.2011 15:23, schrieb carol white:

Hello,
It might be an easy question but if you have many variables to fit in the lm function, how do you take all without specifying var1+var2+...+var2100 in the terms parameter in response ~ terms?

Cheers,

Carol