Mon, 18 Sep 2000 16:05:30 +1000
Here are some further possibilities:
1. An equivalent of S-PLUS plot.gam(), which I
would call plot.terms().
[predict(.., type="terms") provides the
information that is plotted.]
2. An option in print.lm() summary.lm() etc.
that would relate the number of decimal
places shown for coefficients etc. to the SE,
e.g. a precision equivalent to the 2nd d.p.
of the SE.
(This may be hard to get right!)
3. Label output where factors appear so that it
is clear what parameterisation has been used.
(Peter Dalgaard suggested a scheme some time
ago that seemed to me eminently sensible.)
4. As an interim measure, until lattice arrives,
implement simplified versions of trellis
bwplot, qqmath, etc that allow two conditioning
factors. This can be done as fairly easy
adaptations of coplot(). So that it is not
repeated in each individual function, the
coplot() code that decodes the graphics formula
should come out into a separate function.
It needs slight modification to allow formulae
such as ~x|a+b.
(I have a version of such a function. Adding
the code for e.g. a primitive version of
bwplot is straightforward. I will do it myself
shortly if there are not other offers.)
5. I have been investigating a function that has
one list of information for each panel in a
trellis type layout, then allowing the user to p
rovide a panel function that operates on the list
elements. For example, one may want graphs with
different numbers of x-values to be superimposed
in the same panel. Or a theoretical curve or plot
of simulated data may be specified by a small
number of parameters that are specific to each
(I have a primitive version of such a function.)
John Maindonald email : email@example.com
Statistical Consulting Unit, phone : (6249)3998
c/o CMA, SMS, fax : (6249)5549
John Dedman Mathematical Sciences Building
Australian National University
Canberra ACT 0200
r-devel mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !) To: firstname.lastname@example.org