Re: st: Why is the legend in xtline limited to 2 to 15?

[Nick Cox <njcoxstata@gmail.com>]

Excellent! I imagine many people would want to hear much more.
Nick
njcoxstata@gmail.com


On 12 March 2014 11:20, Austin Nichols <austinnichols@gmail.com> wrote:
> re: " interactive graphic in which
> names pop up on demand and the curious reader can interrogate at will.
> I leave this on one side because you can't do in Stata."
>
> It's relatively easy to write out some JavaScript from Stata to go
> with a series of graphs, e.g. the graphics in
> http://blog.metrotrends.org/2014/03/good-dcs-schools/
> are straight out of Stata, which also writes the content of the drop-down menu.
>
>
>
> On Wed, Mar 12, 2014 at 7:05 AM, Nick Cox <njcoxstata@gmail.com> wrote:
>> Peeling off the last question: Don't use a legend for one item. Seeing
>> that reminds me of a certain spreadsheet application, which if I
>> remember correctly uses by default the legend "Series 1" for one
>> variable; it doesn't even say "G", or whatever, which in its own terms
>> would mean something. Use a title or (usually better) a subtitle for
>> that information; you can just add it yourself.
>>
>> On the main question:
>>
>> I am guessing here, but have a look at -help pstyle- for a clue. By
>> default there are 15 different plotstyles, so I imagine that if you
>> don't spell out 17 or 20 different ways of showing 17 or 20 different
>> countries, -twoway- is just going to cycle around the same
>> possibililities from 16 onwards.  I don't think I've ever tried it.
>>
>> I am impressed at your optimism that this could work well at all. With
>> 17 or 20 countries,
>>
>> a. Clearly the country names mean something and you would prefer
>> seeing them, but a legend with that many entries will take up a large
>> fraction of your real estate.
>>
>> b. The plot itself is all too likely to look like a tangled mess.
>>
>> This is a standard problem, and a difficult one, and all solutions are
>> wrong from some point of view, but let's try some positives:
>>
>> 0. There's a serious argument, which is hardly novel, that the only
>> way to do this properly is through an interactive graphic in which
>> names pop up on demand and the curious reader can interrogate at will.
>> I leave this on one side because you can't do in Stata. Whether it
>> would be the answer for Nirina I can't tell.
>>
>> 1. Separating the series, one to a panel, might be tolerable. Names
>> would be shown better, but clearly you lose out on how easily you can
>> compare the series. Naturally that's the default for -xtline-. (I
>> guess that Nirina is using the -overlay- option.)
>>
>> 2. There is a compromise in which every panel contains all the series,
>> best shown as a backdrop, but each panel highlights just one series.
>> The main idea occurs in many places, but two that spring to mind are
>> http://www.stata-journal.com/sjpdf.html?articlenum=gr0046 and
>> http://www.stata-journal.com/sjpdf.html?articlenum=gr0040 The links
>> lead to .pdf accessible to all.
>>
>> 3. Do you need all the countries to be identifiable? Sometimes what
>> works reasonably is to show all and just to identify only a few
>> countries of greatest interest. We don't know what the research is
>> here, but it is rare in my experience that every country really needs
>> individual identification and comment.
>>
>> Nick
>> njcoxstata@gmail.com
>>
>>
>> On 12 March 2014 08:56, Nirina F <fstata@gmail.com> wrote:
>>
>>> I have panel data and I am graphing a simple line graph through
>>> years for 17 and then 20 countries with xtline. It looks like the
>>> legend is only limited to 15 countries so then it reports to me two
>>> countries under the same legend such as this ---------USA/Luxembourg
>>> and ______China/Uruguay when I graph 17 countries and then five
>>> countries under the same legend for 20 countries. do you know how to
>>> expand it so it is not confusing like this?
>>>
>>> Also, when I am just graphing one country
>>> xtline yvar if Country=="USA", overlay then I do not get the legend
>>> that that single line is for USA and I do not see any option to add
>>> legend.
>> *
>> *   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/
> *
> *   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/
*
*   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/