Sorry, but I think your supposed justification for
upwards and downwards designs misses my comment
completely. For example, an upwards and upwards
design is based on exactly the principle you
cite, so common support does not dictate
graphical design.
My comment is based on graphical statistics literature
focusing on what kinds of comparison are easy
and effective. "The elements of graphing data"
by William S. Cleveland (full reference in
the Stata manuals) leads to some literature.
Nick
[email protected]
Hans J. Baumgartner
> That?s exactly the reason why propensity scores should be
> shown in the
> ?upwards and downwards? design, since the graph shall show that both
> groups do have a common support area.
>
> I?ll check the reference anyway.
> > Upwards and downwards designs appear popular
> > for no good reason. In effect the reader is
> > expected to be able to pick up one length,
> > transfer it and superimpose it, all in one's
> > head, upon another length. Why this should be easier
> > or more effective than comparing juxtaposed lengths
> > beats me.
> >
> > The same issue arises with left and right
> > (side-by-side) designs such as population
> > pyramids.
> >
> > In each case, small and subtle differences
> > could easily be of interest or importance.
> >
> > The problem with histograms is naturally the
> > loss of detail produced by binning. Often
> > this is unimportant but frequently a researcher
> > does want to be sure that is so.
> >
> > To compare two sets of values qua distributions,
> > -qqplot- is pretty nearly an optimal plot. To
> > compare them as paired values, there are several
> > good methods. A fairly lengthy discussion with
> > references is given in
> >
> > Graphing agreement and disagreement.
> > Stata Journal 4(3): 329--349 (2004)
*
* 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/
