RE: st: graphs

 From "Nick Cox" To Subject RE: st: graphs Date Mon, 15 Dec 2008 16:39:56 -0000

```You are correct. -mrunning- does not try to give a direct plot of the
data; it tries to do something that in my experience is often as or more
more useful for your kind of problem, which is why I suggested it.

I find your wording very odd. It seems more natural to me to think of
the response (DV, if you will) varying with respect to one or more
predictors. If that is not how you think about it, why do you consider
it to be the DV?

However, that is a minor detail. As I said earlier, and your comment
here strengthens it, your problem sounds four-dimensional to me. I don't
think S-plus or R can oblige.

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

Victor, Jennifer Nicoll

Thanks Nick and David.  I played around with the mrunning command a bit.
Unfortunately, it does not seem able to show how the data vary across
time for two variables with respect to the DV.  I had hoped to find a
way to put this all on one graph.  It seems David suggestion of doing
this in R is perhaps my only path.

To give you an example of what I'm trying to accomplish.... I have a
dependent variable that is a measure of policy specialization for a
panel of members of Congress across 34 years.  I want to show how their
rates of specialization vary depending on (1) whether they sought higher
office AND (2) the relative difference in demographics between their
district and their state (such as the % of blacks).  The only way I can
think to show all this on one graph is with a 3-D graph.  I guess that
means using s-plus or R.

Nick Cox

I meant -mrunning- from the Stata Journal, as David points out.

Nick Cox

David I think meant not in official Stata.

Jennifer's problem sounds, from another point of view, four-dimensional.

Even if it is, -mrunning- from SSC offers another alternative.

David Airey

No, not in official data.

Often there are two dimension reductions that serve just as well. The -
vibl- package is one example of useful two dimensional plots, where a
3 dimensional plot is a natural choice,
(http://www.ats.ucla.edu/stat/Stata/seminars/stata_vibl/vibl.pdf
).

But anyway, you have a free alternative in several R packages, such as:

<http://www.statmethods.net/graphs/scatterplot.html>

Commercial packages like JMP 7 or 8 do this kind of plot well. Also
Data Desk 7, but development has stopped on this package.

On Dec 12, 2008, at 4:02 PM, Victor, Jennifer Nicoll wrote:

> Can stata make a three-dimensional graph?  I want to show how two
> variables (one dichotomous and one continuous) change over time with
> respect to a dependent variable.  I can envision something three-
> dimensional getting the point across but haven't found anything in
> the manuals to lead me in that direction.  Is it possible?

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/
```