Re: st: Re: Odds ratio graph - selecting order of bars

[Nick Cox <njcoxstata@gmail.com>]

I share many of David's prejudices about graphics,  but not all.
Stacked or divided bars are, in my view, also over-used in total, but
whenever the primary aim is to show variations in totals and a
secondary aim is to show components, it is dubious whether equivalent
charts based on separated bars or dot charts necessarily work better.

Statistical graphics is, like many fields, an odd mixture of sound
logic, arbitrary convention and occasional fallacy. I've experimented
with groups of students (typical age 20) showing them bar charts and
dot charts (in the sense of -graph dot-) of the same data, and there
is an overwhelming preference for bar charts. Repeatedly the
justification is just familiarity. These students prefer a graph form
they have known for  a decade or more and are unwilling to go for
something cleaner and simpler. Naturally, the answer is to keep on
pushing.

In terms of David's (excellent) references, it is too often forgotten,
or not appreciated, that Edward Tufte wrote several books before the
one cited here. Cleveland's 1985 book went to a second edition in
1994; he self-published from Hobart Press. Robbins' book has just been
reissued in a cheaper form by Chart House.

Nick
njcoxstata@gmail.com


On 10 July 2013 12:42, David Hoaglin <dchoaglin@gmail.com> wrote:
> Hi, Mike.
>
> Nick Cox has suggested a good way to learn about Stata graphics.  The
> tools are impressive, and I continue to be surprised at the displays
> that creative and skillful users have developed.
>
> Clever programming, however, does not necessarily produce effective
> displays.  Work on statistical graphics has a substantial literature
> and has led to a good understanding (supported by empirical studies)
> of which techniques are effective and which are not.  I won't try to
> give a bibliography, but Ed Tufte's four books (starting with his 1983
> one) and Cleveland (1985) are good places to start.  Howard Wainer has
> written a stream of interesting books (e.g., 2009), and he has a
> column in CHANCE magazine.
>
> Nick commented that my comments about poor graphical displays on that
> site were not specific. When I wrote that comment, I did not have time
> to go into detail.  But also, after scrolling through that site, I did
> not know where to start.  Many of its examples illustrate clever
> programming and poor displays (though one could put the programming to
> good use).  For example, I noticed a variety of stacked and divided
> bar charts.  Because of its shortcomings that type of display should
> generally be avoided.  (A key shortcoming of stacked bar graphs is
> that categories other than the bottom one are made difficult to
> compare among groups by not having a common baseline.)  The book by
> Naomi Robbins (2005) gives further explanation and examples of better
> displays.
>
> My advice: Choose graphs that are effective for displaying the type of
> data you have, and then learn how to program them in Stata.  If
> someone hasn't already created the graph you need, Stata probably has
> the tools to do it.  If everyone in your field uses a particular type
> of graph, don't assume that that graph is effective.  Old habits die
> hard.  For example, bars are heavily overused, and some journals
> contain many figures that use "dynamite-plunger plots" (discussed by
> Freeman et al. 2008 in the chapter How to Display Data Badly).
>
> David Hoaglin
>
>
> Cleveland, William S. (1985).  The Elements of Graphing Data.
> Monterey, CA: Wadsworth.
>
> Freeman, Jenny V., Walters, Stephen J., and Campbell, Michael J.
> (2008).  How to Display Data.  BMJ Books.
>
> Robbins, Naomi B. (2005).  Creating More Effective Graphs.  Hoboken, NJ: Wiley.
>
> Tufte, Edward (1983).  The Visual Display of Quantitative Information.
>  Cheshire, CT: Graphics Press (2nd ed, 2001).
>
> Wainer, Howard (2009).  Picturing the Uncertain World: How to
> Understand, Communicate, and Control Uncertainty through Graphical
> Display.  Princeton, NJ: Princeton University Press.
>
> David Hoaglin
>
> On Tue, Jul 9, 2013 at 8:18 PM, Michael Stewart
> <michaelstewartresearch@gmail.com> wrote:
>> HI ,
>> Would David/Nick or any statalist members recommend sources to learn
>> stata graphics , especially involving some programming like in
>> http://www.survey-design.com.au/Stata%20Graphs.html
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/