# RE: st: scatter with aweight - consistent sizing across subsets of observations

 From "Nick Cox" To Subject RE: st: scatter with aweight - consistent sizing across subsets of observations Date Thu, 26 Mar 2009 19:11:55 -0000

```To Friedrich's specifics, let me add a reference to a  more general comment:

<http://www.stata.com/statalist/archive/2006-06/msg00291.html>

My dislike of these graphs shouldn't matter to anyone but a few close friends who might be interested in my tastes. But there is, I think, a more serious general point. People often want these graphs to show symbol sizes proportional to some specified variable, but that is only rarely, if ever, either practicable or produced. At most, expect a monotonic relation within any graph, or across graphs with exactly the same ranges.

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

Friedrich Huebler

You can find the solution to your problem in this post:

http://www.stata.com/statalist/archive/2008-08/msg00987.html

The example uses overlaid scatterplots but the principle applies to
plots that are drawn separately. Try the code in the post and add one
new command at the end:

. twoway scatter y x [w=weight], by(group) name(D) legend(off)

A more complicated example is shown here:

http://www.stata.com/statalist/archive/2008-08/msg00982.html

On Thu, Mar 26, 2009 at 12:22 PM, Dan Weitzenfeld
<dan.weitzenfeld@emsense.com> wrote:

> I am confused as to how exactly Stata determines what will be the
> largest and smallest marker size when using -scatter- with aweights.
> My goal is to create a series of aweighted scatterplots, each one on a
> subset of my data, with the relative marker sizing consistent across
> plots.  That is, if a point in my first graph and a point in my second
> graph are both weighted 100, I'd like them to be the same size,
> regardless of what other points are plotted on each.
>
> My idea for how to do this is to create two fake datapoints, one equal
> to the universal minimum and one equal to the universal maximum of the
> weights, and to include these fake datapoints in each plot.  Since I
> am copying the graphs into PowerPoint and ungrouping them anyway, I
> can just delete these two fake points.

*
*   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/
```