Re: Re: st: catplot problem

[n j cox <n.j.cox@durham.ac.uk>]

Anyone trying to follow this thread can find the previous material here:

<http://www.hsph.harvard.edu/cgi-bin/lwgate/STATALIST/archives/statalist.0805/date/article-230.html>

Tunga thinks it a "deficiency" that my -catplot- program, downloadable 
from SSC, when faced with a subset of observations with categories 1 4 5 
6 7 8 9 10 does not notice that elsewhere in the data there are 
categories 2 3, which he wants to show on a graph. He wants a "better 
developed" command which does notice that.

In general, Stata graph commands do not do that, even optionally.

To repeat, -catplot- in particular was not designed to work like that, 
even optionally. Tunga should feel free to write a program that does 
what he wants, although ensuring that bars of zero height are drawn for 
categories that don't exist, but might have existed, in the observations 
being used  may prove trickier than he thinks for any approach based on 
-graph bar-. (That is straightforward if you switch to -graph twoway bar-.)

However, that is all quite unnecessary, as -histogram- (a command, not a 
function) should prove capable of doing what he wants, or at least I 
have seen no evidence to the contrary. -histogram- does support 
categorical variables, with the -discrete- option that Tunga is already 
aware of (so I still don't understand why he keeps asking this 
question). If a categorical variable is string, map to numeric first 
using -encode-. Then show all the gaps you want using -xscale()- and 
-xlabel()-. That can be automated too:

Simple example:

sysuse auto
levelsof rep78, local(levels)
histogram rep78 if foreign, xla(`levels') discrete freq barw(0.8)

Otherwise put, I do see wanting -catplot- to behave like -histogram- in 
Tunga's circumstances as perfectly reasonable, but the answer is just to 
use -histogram-.

If Tunga can identify a specific kind of graph in this territory that it 
seems neither -catplot- nor -histogram- can produce, then I or others on 
the list will  be happy to try to make suggestions -- or write new code 
if needed.

Nick
n.j.cox@durham.ac.uk

Tunga Kantarci

Dear Nick,

Thanks for your reply.
I am not sure if I should have called it a bug, but I also don't know what
to call it. For me it is a deficiency.
I would consider catplot to have been better developed if it was able
determine missing values over which the category is defined, i.e. 2 and 3 in
the example.
Because currently it is just not able to match with the values of the
variable of interest over which the categorizations are intended.
Or the histogram function could be developed further to allow
categorizations.


*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/