rmh comments to John Fox

Happy New Year and I am back to working with Rcmdr.  Thanks again for
the program.  I didn't teach the relevant course in Fall 2006 and I am
teaching it again in Spring 2007.  Burt did teach it in the Fall and I
am trying to get his wish list in place before classes start next
week.  Some of them are included here.  Much off this version is catching
up with the changes you have installed since the summer.

You can look at Rcmdr.HH_1.4 on my web page.  I have not yet submitted
it to CRAN.
http://astro.ocis.temple.edu/~rmh/Rcmdr.HH

This is based on R-2.4.1 and R Commander Version 1.2-6


Here is a list of things in the order that I found them.
Some are things I wrote, some are recommended changes to Rcmdr,
some are comments, some are wish list.

1. I see that your data menu now reads Excel datasets directly, in
   addition to SPSS and Minitab files.  On my wish list is SAS
   datasets.


2. I clicked
   Data/data in packages/read data/datasets/iris3
Rcmdr made a copy of the iris3 data in my ".GlobalEnv" and
gave the error message
ERROR: iris3 is not a data frame and cannot be attached.

a. is this an "ERROR"?  The dataset was found and is now in my environment.
Therefore I think it should be a "WARNING".

b. I think a better response from Rcmdr would be to convert it to a
data.frame and issue a warning about the conversion.


3. typo: missing word "are"
ERROR: There no factors in the active data set.
83175 Oct  4 20:58 utilities.R


4. print.summary.regsubsets in leaps doesn't follow the rules about
the behavior of print methods.  I sent revised
print.summary.regsubsets to Tomham Lumley.  I don't use this function
since I have print.summary.HH.regsubsets instead.

"print.summary.regsubsets" <-
function (x, ..., quote=FALSE) {
  print(x$obj)
  print(x$outmat, quote=quote)
  invisible(x)
}


5. Not quite the right word in graphs-menu.R, lines 939 and 943.
You are using the word "range" to mean
     max(x)-min(x)
To me "range" means the interval
    (min(x), max(x))
Therefore I don't like the use of "Same range" as a synonym for the
trellis term "sliced".  I prefer something more descriptive,
something like "same number of units per inch".


6. stack() doesn't take factors.  I think this is a design error in
stack().  I don't plan to fix it this week.  reshape() might be a
better choice for a menu item.  reshape has more options to place into
the menu.


7. Reading Excel datasets prints an invalid statement (or perhaps just
an incomplete set, since the channel is not defined) in the log.  I
tried to repeat the command by clicking the generated statement in the log
and it gave me an error.  I am not going to solve this now.

ERROR:  syntax error in "Xr13.119 <- sqlQuery(channel = 3, select * from [Sheet1$]"


8. The MS Open dialog box for Excel does not remember the directory
that I opened in the previous time in the same Rcmdr session.  For
classroom demonstrations and for students doing homework, this would
be important.  The sheet dialog which replaces the open dialog is very
tiny and easily overlooked.  I would prefer the sheet dialog to
overlay the open dialog so that I will see it and visually connect it
with the current command.


9. The stack menu item proposes the default name "StackedData".  I
would always like the name OldNames, as in X13.119s <- stack(X13.119[,
c("X.100.Limit","X.3000.Limit")]) Is there a way (Rcmdr option
perhaps?) to change the default to base the proposed name on the old
name?



10. confint can have very long row names, leading to default graphs
that can't be read.

From the Rcmdr menu, load the barley data from lattice, then click
statistics/means/oneway ANOVA
click variety and yield and pairwise.

Most of the labels are clipped on the left.
This example needs a manually entered
  par(oma=c(0,9,00))
in order to read the graph.

Can we put some control of oma onto the menu?


11. scatterplot.matrix
Now that you have included my density, I am using your function.  I
still need my own menu into it because I consider the row1attop=FALSE
argument to be critical.  If you make it a user option on the menu,
even initialized to TRUE, then I will no longer need to maintain a
separate menu.


12. 3d scatterplot

There were two items in my version that you dropped.

The option to open a new 3d window is important.  When I rotate these
graphs I like to be able to compare two or more graphs.  The relevant
command is
   rgl.open()
and this can be entered manually.  I agree that it doesn't have to be
on the "3D graph" menu.  I would like it as an item on the main
"Graphs" menu.

The other item is to plot a plane that is not the
least-squares plane.  This is pedagogically useful.
It allows us to show what "least squares" means.
I think some kind of syntax to permit this option is important.
I agree that the quick and dirty multiplier is not optimal.


14. setOption in Commander in commander.R is very subtly broken.
     current[[option]]
needs to be replaced by
     current[option][[1]]
in two places.
Reason, $ and [[ both use partial matching and [ does not.
Thus "etc" and "etcMenu" are seen as identical when
options()$Rcmdr$etcMenu exists and options()$Rcmdr$etc is NULL.
I worked around this by assigning both.  This is safe at the moment
because I am probably the only user of this feature.
