Skip to content

OSX R Gui Data editor

3 messages · John Walker, Peter Cowan, Byron Ellis

Original
John Walker
# 
G'day

I just tried to introduce a naive user to R via doing a t-test on her
data using Rcmdr on OS X. Now I'm not sure if this is the right mailing
list but I must say that  the introduction was a complete and total
failure because of the OSX data editor that is called by Rcmdr. I think
this is the one that comes with the R GUI. We tried to enter her data
into R and I couldn't do it using the OS X data editor. When I had to go
to the command line she was clearly initialy put off by the poor editor
and by the need for the command line to fix it.  As she said, "how can I
trust a stats program with my important data if they can't get data
entry working?". She promptly went back to Instat

I know that the criticism she voiced isn't valid and that some learning
is required for any program, but I think she has a point. The data
editor as it stands is a failure. The idea that you can add a column but
not change the variable name and that you have to double click on each
cell to enter data is crazy. It is simply a pretty looking window with
less functionality than the old x11 interface.

I'm aware that the current model is "prepare your data elsewhere and
import it into R" but this is an absurd stance. It works for large data
sets that statistical experts deal with but then the  casual t-test
requires several programs and an import step. A silly approach
especially when the data sets are small (about 10-20 entries) and all
the user wants to do is a simple t-test.

If braoder adoption of R is an aim then the OS X data editor needs to be
at least as functional as the X11 one. The ability to double click to
change variable names and right click to change the variable type (or
menu entries to do this) is important if not essential. The ability to
tab to the next data entry slot is also a simple but important function
that needs to be included.  Right now it is a barrier to new OS X users
who want to try out R using a simple t-test on a small data set.


Regards


John Walker
Peter Cowan
# 
John,

I think you may be mixing up the R package Rcmdr [^1] with the Mac OS
X R gui.  I've haven't used Rcmdr in years so I cannot comment on your
issue, but I think you were commenting on the Mac gui.

Yes, I agree that the edit() function leaves a bit to be desired and
using it as the primary data entry point would be painful.  Your
suggestions for improvement would make it more user friendly.
However, the gui currently implements the edit.data.frame() function
which is a part of R (as opposed to the gui) and thus works on other
platforms, it appears to allow editing of row names, but not column
names.  I don't believe it has ever been intended as the primary means
of getting data into R.  It is, however, useful for fixing the
occasional incorrect value in a dataframe.

Perhaps what you are looking for is something along the lines of the
xlsReadWrite package, which purports to read and write Excel files
(though having excel save as csv usually works for me).

If your colleague isn't interesting in learning to use a command line
program, then I suspect that R is not the correct choice for her.
Developing guis is a lot of work, and actually I think that the Mac R
gui is impressive and quite nice to work in compared to the windows
gui.

Lastly, please keep in mind that R is an open source project.  The
people who develop have kindly license their work so that we can all,
use it, learn from it, and improve it to match our needs if we desire.

Peter

[1] <http://socserv.socsci.mcmaster.ca/jfox/Misc/Rcmdr>
On Mon, Mar 10, 2008 at 1:11 PM, John Walker <john.s.walker at uchsc.edu> wrote:
Byron Ellis
# 
Your colleague's reaction is completely correct. If all you need to do
is hand enter 20 data points and run a t-test then even Instat is
probably overkill and R is definitely overkill. Hell, I'm pretty sure
the last graphing calculator I owned could do various sorts of
t-tests. The ones kids have these days can probably do glm.

R is mostly aimed at people whose data is already in electronic form
so the data entry mechanisms simply don't get used very much
(certainly not by me, the idea of hand entering data mostly strikes me
as a good way to introduce errors). Think of R like a big industrial
CNC mill or something. It's big, clunky and kind of a pain, but if you
need to make a set of intricate wheel rims to very high tolerances
there is nothing on this earth that can match it. However, not so
great for paper snowflakes, unless you've been using it forever in
which case its mostly out of habit... and you've lost your scissors.

Now, would it be nice if there was a full-blown spreadsheet interface
built into R? Hell, yes. But, unless someone is volunteering to write
it, I'm more likely to discover that my alarm clock is made of pound
cake.
On Mon, Mar 10, 2008 at 1:11 PM, John Walker <john.s.walker at uchsc.edu> wrote: