Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: RE: STATA Wish List

From   "Nick Cox" <[email protected]>
To   <[email protected]>
Subject   st: RE: STATA Wish List
Date   Wed, 20 Oct 2004 12:17:40 +0100

Comments interspersed. In general, the most effective way 
to get precisely what you want is to program it yourself! 

[email protected] 

[email protected]

> If anyone can point me to ado files that can accomplish the following
> various tasks (or have the described features), I'd be most 
> grateful. I
> suspect that some (for understandable reasons) do not exist - 
> in that event
> this email may at least serve as my  wish list (pretending my 
> wishes count) for the programmers at STATA.


> *	A variable information screen which is similar in 
> appearance to a
> spreadsheet (with scroll facilities) - each row provides 
> information on one
> variable - name, variable and value labels. This is great for 
> pinpointing
> variables of interest in a dataset containing a thousand 
> variables which you are not familiar with. 

Sounds as if it would be fairly popular. How much better would 
it be than smart use of -describe- and -ds-? This could 
be simulated at least crudely by a user programmer, although 
only as a matter of report, not as offering an interface 
for changing anything. 

> *	The ability to have more than one syntax (STATA do 
> file) window open
> at any one time (and no character limits in each one). I like 
> to have one
> syntax file holding data manipulation commands and another 
> file holding the actual 'analysis' commands.

Sounds as if it would be fairly popular. 

> *	Ability to produce column and row percents in n-way contingency
> tables. I know I could use STATA's  "Tabulate" command with 
> the "by" command
> to achieve this but the presentation is much better in STATA's "Table"
> command but it does not have these percent options.

I think there would be widespread agreement that tabulation facilities
in Stata are slightly confusing at present, and that middle term 
at least, we might expect some tidying up and extension from 
StataCorp. For the moment, -groups- from SSC does some of what may 
want. I think it's also true that there are so many possible desired 
kinds of tables that being prepared to write a few lines computing 
the percents yourself gives considerable extra flexibility. 

> *	A command similar in spirit to STATA's "tab2" command 
> but instead of
> showing every possible two-way tabulation of the variables 
> specified in
> varlist you specify a varlist to act as rows (e.g.. 3 variables) and a
> separate varlist to act as columns (e.g. 4 variables) and the 
> command then
> produces every two-way association of the vars in rows with 
> those in columns (e.g. 3*4=12).

A long time ago I wrote a command -biv- which purports to 
do this (STB-45). I don't think it ever caught on, even with 
myself, partly 
for some good reasons. I think the main answer here is that the flexibility 
you really want is best provided not by a single command, whose
syntax would be impossibly cumbersome for what you would probably
expect of it, but by getting familiar with -foreach- and -forvalues-.  
I guess that you should not expect your overall control command
to know about what the command it is controlling is doing, yet a user might 
keep slipping into that expectation. With -foreach- and -forvalues- 
that problem is solved as the programmer should know what the controlled 
command is doing. 

> *	Ability to quickly copy over selected output straight 
> to Microsoft
> Word which looks as good as it does in the output window. I 
> know you can
> achieve this by running STATA's "translate" command but a 
> quicker way would
> be preferred. A two way output screen similar in appearence 
> to SPSS's that
> allows users to effortlessly copy over selected output to 
> Word would be
> ideal (but I think that's asking a bit much).

No doubt popular. There's possibly an answer involving more
explicit support for HTML within Stata. Expecting MS Word 
to support SMCL is no doubt unrealistic. 
> Please do not think I am trying to denigrate STATA in any way. For all
> "serious analytical"  work I use STATA. Such work by its very 
> nature is
> focused on relatively small number of variables where the 
> analysis output
> (however advanced) is not copious and is not impacted by any 
> of the lack of
> features mentioned above. Unfortunately in my commercial 
> environment I'm
> often called upon to produce lots of mind numbing contingency 
> tables and it
> is solely in this regard that SPSS has the advantage.
> On a purely separate topic in STATA 7 there was a command 
> "graph oneway"
> which with the addition of the sub option "box" plotted both 
> a boxplot and
> the associated raw points together but could not plot a 
> scaled axis (other
> than listing the axis endpoints) . Nick Cox wrote a nice ado 
> "onewplot2" ado
> which could plot the scaled axis but did not show the box 
> plot. Does anyone
> know if there is a command that can plot the boxplot together 
> with the raw
> output on a scaled axis? I have STATA 8 but cannot locate any 
> such graphs.

I've lost track of whatever -onewplot2- was. -onewplot- for Stata <8 
was superseded by -onewayplot- for Stata 8 in summer 2003, and 
there are examples in the help of how to add symbols for quartiles. 

There's an example of a hybrid oneway and box plot in Vince Wiggins' 
London and Boston graphics talk, which I think is attributed to me. 
No doubt that could be wrapped up as a single command, so long as 
you didn't expect too many handles. 

Doing it properly so that you have all the control of box plots 
you have with -graph box- and all the control of oneway plots 
you might want sounds non-trivial, even for StataCorp. 

*   For searches and help try:

© Copyright 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index