R devel - Mar 2005 - improved pairs.formula?

Dear all,

I would like to suggest changing the pairs.formula command such that a
command like

        pairs(GNP ~ . - Year - GNP.deflator, longley)

would behave in a similar fashion as 

        lm(GNP ~ . - Year - GNP.deflator, longley)

i.e., make a pairwise scatterplot of GNP and all other variables in
the (longley) dataframe except for Year and GNP.deflator.  The above
command, with the current version of pairs.formula, produces a
pairwise scatterplot of all variables in the (longley) dataframe.

After some tinkering around, I came up with the following replacement
function of pairs.formula which seems to do the job:

    pairs.formula <- 
    function (formula, data = NULL, ..., subset, na.action = stats::na.pass) 
    {
      m <- match.call(expand.dots = FALSE)
      if (is.matrix(eval(m$data, parent.frame()))) 
        m$data <- as.data.frame(data)
      m$... <- NULL
      m$na.action <- stats::na.pass
      m[[1]] <- as.name("model.frame")
      mf <- eval(m, parent.frame())
    
      mt <- attr(mf, "terms")
      tmp <- attr(mt, "factors")
      ind <- apply(tmp, 1, max)
      sv <- rownames(tmp)[ind>0]
      ind <- match(sv,
                   sapply(attr(mt, "variables"), deparse,
width.cutoff=500)[-1])
      if( (tt <- attr(mt, "response")) != 0 ){
        ind <- c(tt, ind)
      }
    
      pairs(mf[,ind], ...)
    }

Would you please kindly consider replacing the current pairs.formula
function (at the top of the file src/library/graphics/R/pairs.R) with
the above function?

Best wishes,

        Berwin

========================== Full address ===========================Berwin A
Turlach                      Tel.: +61 (8) 6488 3338 (secr)
School of Mathematics and Statistics        +61 (8) 6488 3383 (self)      
The University of Western Australia   FAX : +61 (8) 6488 1028
35 Stirling Highway                   
Crawley WA 6009                e-mail: berwin@maths.uwa.edu.au
Australia                        http://www.maths.uwa.edu.au/~berwin

On Tue, 29 Mar 2005, Berwin A Turlach wrote:
> Dear all,
>
> I would like to suggest changing the pairs.formula command such that a
> command like
>
>        pairs(GNP ~ . - Year - GNP.deflator, longley)
>
> would behave in a similar fashion as
>
>        lm(GNP ~ . - Year - GNP.deflator, longley)
>
> i.e., make a pairwise scatterplot of GNP and all other variables in
> the (longley) dataframe except for Year and GNP.deflator.  The above
> command, with the current version of pairs.formula, produces a
> pairwise scatterplot of all variables in the (longley) dataframe.
>
> After some tinkering around, I came up with the following replacement
> function of pairs.formula which seems to do the job:
>
>    pairs.formula <-
>    function (formula, data = NULL, ..., subset, na.action = stats::na.pass)
>    {
>      m <- match.call(expand.dots = FALSE)
>      if (is.matrix(eval(m$data, parent.frame())))
>        m$data <- as.data.frame(data)
>      m$... <- NULL
>      m$na.action <- stats::na.pass
>      m[[1]] <- as.name("model.frame")
>      mf <- eval(m, parent.frame())
>
>      mt <- attr(mf, "terms")
>      tmp <- attr(mt, "factors")
>      ind <- apply(tmp, 1, max)
>      sv <- rownames(tmp)[ind>0]
>      ind <- match(sv,
>                   sapply(attr(mt, "variables"), deparse,
width.cutoff=500)[-1])
>      if( (tt <- attr(mt, "response")) != 0 ){
>        ind <- c(tt, ind)
>      }
>
>      pairs(mf[,ind], ...)
>    }
>
> Would you please kindly consider replacing the current pairs.formula
> function (at the top of the file src/library/graphics/R/pairs.R) with
> the above function?
Perhaps you could explain what precisely 'the job' is and why you chose 
such an unusual piece of code to do it?  (E.g. what is the prescription 
for the ordering of terms, and why do you think the rownames of the 
factors and the variables might be in different orders?  They are set the 
same in the C code.)

BTW, the help page specifically warns against a formula of the type you 
specified.  Why do you want to allow a response?  Currently only '+' is 
documented to work.

-- 
Brian D. Ripley,                  ripley@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