# st: RE: Confidence Intervals plot

 From "Nick Cox" To Subject st: RE: Confidence Intervals plot Date Tue, 13 Sep 2005 22:02:05 +0100

```Questions of the form "I tried such-and-such
a program, but it didn't produce what I want"
are pretty difficult to answer, especially
if you are fuzzy on what you want.

The programmer is inclined to say, "Well,
feel free to write your own program then!"

-ciplot- is a program on SSC I wrote.

Your reference to -ecplot- I take to mean
one to -eclplot- by Roger Newson.

-ciplot- is a wrapper for -ci-, as its help
explains. I am not clear, as you do not say,
whether you would expect -ci- to produce
your confidence intervals. But if you don't
then -eclplot- should be fine for your purpose.

In general, confidence intervals plots

* the way you like them *

are easy. What is difficult for a
programmer is to prepare for all the ways that
people like them. In your case, you would need
to

calculate your quintile groups

calculate your confidence intervals

decide on how to represent estimates, e.g.
-scatter- or -twoway bar-

decide on how to represent limits, e.g.
-twoway rcap-.

Then a typical plot requires an overlay

twoway ... || ...

Further advice would require a more precise
question.

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

Xavi Barber

> I have two variables A and B (continuous), and C (Yes o No variable)
>
> I'd like to have a plot with the Confidence Intervals plot for A in
> quintiles of B by C
>
> I tried  a lot of plots commands, probably I'm a very bad
> graphs user, but
> I didn't find it.
> ciplot  and variants
> ecplot
> ....

*
*   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/
```