Re: st: create pretty charts

[Nick Cox <njcoxstata@gmail.com>]

On the bottom panel, the dates run (equally spaced)

2/1/08 4/1/08 6/1/08 8/1/08 10/1/08 12/1/08 2/1/09 4/1/09

Anyone else puzzled by this?

On Fri, Sep 7, 2012 at 10:27 AM, Nick Cox <njcoxstata@gmail.com> wrote:
> That's a good exercise for students, to say what is wrong with those
> displays. I refer to
> <http://www.tableausoftware.com/learn/gallery/product-survey-analysis>
>
> 1. State identifiers running alphabetically from CA to WA; that's helpful.
>
> 2. Heavy use of red and green for contrasts: haven't Tableau heard
> that distinguishing red from green is a common  difficulty among
> sighted people?
>
> 3. The projections into the future of interest levels seem to be based
> on fitting quadratics. Interest levels going negative will presumably
> underline emphatically what is dropping off sales-wise. Discuss.
>
> 4. The moving averages seem unsuited to the data, or vice versa.
>
> Still, I must try to be constructive and positive. No doubt Tableau
> nowhere insists on those choices.
>
> I (and other user-programmers interested in graphics) would really
> like to be told about new good kinds of graphs that I can program in
> Stata (and make the code available).
>
> Otherwise, if there is a wishlist I suspect only StataCorp can deliver.
>
> I guess that Maarten has put his finger on a major point. Graphs that
> you can interact with are clearly of interest and -- if they were in
> Stata -- many of us would be as happy as a three-year-old to play at
> producing and to play with them. But when the play is done, it seems
> that almost all researchers using Stata want mostly static graphs for
> their theses, dissertations, reports, papers, books and presentations.
> A dozen programmer-years developing in the direction of interactive
> graphics (wild guess) is a dozen programmer-years not developing new
> models, and model-handling facilities, which is what people directly
> and indirectly ask for on Statalist week in, week out.
>
> Nick
>
> On Fri, Sep 7, 2012 at 9:37 AM, Maarten Buis <maartenlbuis@gmail.com> wrote:
>> On Thu, Sep 6, 2012 at 11:30 PM, Nick Cox wrote:
>>> Pablo and Miguel (and anybody else) would be doing a useful service if
>>> they could specify exactly what in Tableau is really good and also
>>> exceeds what they can do in Stata. That would give programmers if not
>>> an agenda, then a cogitanda.
>>
>> I googled it, and it seems like a piece of software that allows you to
>> program an interactive graph. Some of the example graphs do look
>> pretty, but designing such infographics is best left to professionals.
>> It reminds me a bit of the time that inkjet printers became available,
>> so "normal" people could now freely choose and print any type of font.
>> The result was a mass of horrible and completely unreadable documents
>> with 16 or more different fonts.
>>
>> Even the example graphs provided by the company that makes the piece
>> of software, which should showcase the best use of the software are
>> not convincing: Most of them are pretty but make the content harder to
>> read than necessary, some of them are downright misleading (worst
>> example is: <http://www.tableausoftware.com/learn/gallery/product-survey-analysis>),
>> and only few are useful (I like a set of linked graphs, such that when
>> you select a subset in one graph that same subset is also highlighted
>> in the other graphs. I think that is a nice way to get a feel for
>> multidimensional problems.).
>>
>> So, my impression is that unless you want to start a four year study
>> in graphical design, you'd best leave software like that alone (and if
>> you do finish such a study you'll probably have learned that there is
>> better software for designing such infographics).
*
*   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/