[R] Odp: Loop counter used in variable

[Petr PIKAL]

Hi

Martin Maechler <maechler at stat.math.ethz.ch> napsal dne 15.12.2009 
09:17:55:

> >>>>> "PP" == Petr PIKAL <petr.pikal at precheza.cz>
> >>>>>     on Tue, 15 Dec 2009 09:03:11 +0100 writes:
> 
>     PP> Hi
>     PP> r-help-bounces at r-project.org napsal dne 15.12.2009 00:25:10:
> 
>     >> 
>     >> Hi All,
>     >> 
>     >> I need to run muliple lm functions.  My independent variables are 
called
>     >> dataset$x1, x2, x3, x4 etc.
>     >> 
>     >> How can I use a loop counter variable to replace the numbers?
>     >> 
>     >> 
>     >> 
>     >> fit1=lm(dataset$y~dataset$x1)
>     >> fit2=lm(dataset$y~dataset$x2)
>     >> fit3=lm(dataset$y~dataset$x3)
>     >> fit4=lm(dataset$y~dataset$x4)
> 
>     PP> I would use list for results and brackets for lm fit. Something 
like 
>     PP> (untested)
> 
>     PP> fit<- vector("list", length=4)
> 
>     PP> for (i in 1:4) {
>     PP> fit[[i]] <- lm(y~dataset[,i], data=dataset)
>     PP> }
> 
> Also, in almost all such cases, 
> where you have many regressions,  you'd rather  want  lm.fit()
> instead of lm(), for efficiency reasons.
> 
> lm.fit() is probably among the much R functions that are less
> known than they should be.
> It *would* be known more if useRs *would* read help pages ...
> ?lm  in this case.

Point taken. I do not use lm in cycles very often and when doing so I did 
not encounter performance problem yet. 

R computation is almost always quicker then me when I try to think about 
the problem to be solved correctly.

And personally I prefer to put everything in nice pictures instead of 
reading output from n fits. Therefore I try to make multipage PDF plots 
which can have more than two dimensions by col, pch and size variation 
with other variables. Than I can list through it to see if there is some 
pattern.

Regards
Petr

> 
> Martin