[RsR] Traditional vs. lattice graphics

[Valentin Todorov]

Thanks a lot. Actually my pragmatic question concerned the plots in rrcov,
which I have implemented as base graphics. Since some of them are quite ugly
(these that are multipanel, comparing classic and robust estimates) I
implemented them with lattice/grid but never dared to release such a mixed
version.



Best regards,

Valentin



----- Original Message ----- 
From: "Nicholas Lewin-Koh" <nikko using hailmail.net>
To: <r-sig-robust using r-project.org>; "Valentin Todorov"
<valentin.todorov using chello.at>
Cc: "Martin Maechler" <maechler using stat.math.ethz.ch>; "Achim Zeileis"
<Achim.Zeileis using wu-wien.ac.at>
Sent: Saturday, December 31, 2005 5:53 PM
Subject: Re: [RsR] Traditional vs. lattice graphics


> Hi,
> In general I agree with Martin and Achim except for one small point.
> If you construct your basic plots using grid, it is very easy to
> incorporate
> them into a lattice framework later, since each plot can sit in a
> viewport.
> Even better would be to use a framework so the plot is constructed as a
> layout
> and a series of additions to the plot area, then it is very easy to
> write panel functions
> for lattice, if conditioning is desired later.
>
> My experience with base graphics has been that I can only go so far, and
> I have
> had to switch to grid in the end. Also, grid seems to integrate pretty
> well with s4
> classes, even though Paul used s3 classes for grid. It is very natural
> to
> construct s4 graphics classes with a slot for a vieport tree, and
> various grobs.
>
> That's my 2c
>
> Cheers and happy new year
>
> Nicholas