R help - Mar 2011 - Newbie-ish question on iteratively applying function to dataframe

Hi,
I am trying to recursively apply a function to a selection of columns
in a dataframe. I've had a look around and from what I have read, I
should be using some version of the apply function, but I'm really
having some headaches with it.

Let me be more specific with an example.

Say I have a data frame similar to the following

A     x     y     z     r1    r2    r3    r4
0.1  0.2  0.1 ...
0.1  0.3 ...
0.2 ...

i.e., a number of columns, each of the same length, and all containing
real numbers. Of these columns, I want to model one variable, say A,
as a function of other variables, say x, y, z, and any one of my r1,
r2, r3, ... variables.

i.e., I want to model
A ~ x + y + z + r1
A ~ x + y + z + r2
....
A ~ x + y + z + rn

But where the number of 'r' variables I will have will be large, and I
don't know the specific number of these variables in advance.

My question first is, how can I select all the columns in a dataframe
that have a heading that matches a string pattern?

And then related to this, what would be the best way of repeatedly
applying my modelling function to the result?

Many thanks for any help for this occasional R armature.

Claus

Hi Claus,

On Tue, Mar 15, 2011 at 9:33 AM, Claus O'Rourke <claus.orourke at
gmail.com> wrote:
> Hi,
> I am trying to recursively apply a function to a selection of columns
> in a dataframe. I've had a look around and from what I have read, I
> should be using some version of the apply function, but I'm really
> having some headaches with it.
I would just do it in a loop (see below)
>
> Let me be more specific with an example.
>
> Say I have a data frame similar to the following
>
> A ? ? x ? ? y ? ? z ? ? r1 ? ?r2 ? ?r3 ? ?r4
> 0.1 ?0.2 ?0.1 ...
> 0.1 ?0.3 ...
> 0.2 ...
>
> i.e., a number of columns, each of the same length, and all containing
> real numbers. Of these columns, I want to model one variable, say A,
> as a function of other variables, say x, y, z, and any one of my r1,
> r2, r3, ... variables.
>
> i.e., I want to model
> A ~ x + y + z + r1
> A ~ x + y + z + r2
> ....
> A ~ x + y + z + rn
>
> But where the number of 'r' variables I will have will be large,
and I
> don't know the specific number of these variables in advance.
>
> My question first is, how can I select all the columns in a dataframe
> that have a heading that matches a string pattern?
?grep
>
> And then related to this, what would be the best way of repeatedly
> applying my modelling function to the result?
Well, I don't know about the "best" way. But why not just

set.seed(21 )
dat <- as.data.frame(matrix(rnorm(100000 ), ncol=100, dimnames=list
(1:1000, c("A", "x", "y", "z",
paste("r", 1:96, sep="" )))))

mods <- list()
for(i in grep("r", names(dat ), value=TRUE)) {
    mods[[i]] <- lm(as.formula(paste("A ~ x + y + z + ", i)),
data=dat )
}

Note that  you should be cautious about making any inferences based on
this kind of method. In the example above 9 r variables are
"significant" at the .05 level, even though the data was generated
"randomly":

sort(sapply(mods, function(x) coef(summary(x))[5, 4]))

Best,
Ista
>
> Many thanks for any help for this occasional R armature.
>
> Claus
>
> ______________________________________________
> R-help at 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.
>


-- 
Ista Zahn
Graduate student
University of Rochester
Department of Clinical and Social Psychology
http://yourpsyche.org