Re: st: Horizontal time line graph of individual with time events

[Nick Cox <njcoxstata@gmail.com>]

Even the more complicated graph you cite looks doable quite nicely in Stata.

For that and your own problem the way ahead is to break it down into
separate tasks for -twoway-. For example, the little boxes on the
cited graph would just be -ms(Sh)- shown big enough to contain the
other symbols. That is the trick behind the first -stripplot- example
given at

http://www.survey-design.com.au/Usergraphs.html

Some technique is discussed at

SJ-8-2  gr0034  . . . . . . . . . .  Speaking Stata: Between tables and graphs
        (help labmask, seqvar if installed) . . . . . . . . . . . .  N. J. Cox
        Q2/08   SJ 8(2):269--289
        outlines techniques for producing table-like graphs

but for the most part it's probably easier to go straight at it as an
application of -twoway rspike || scatter || scatter || scatter-.

For your cited graph I would create a vertical axis variable that just
goes 1, 2, 3, and then put age and sex into value labels like "2 m" "5
m". The reader would not see the y values, just the value labels.

Remember that whenever one is thinking table style (i.e. lowest row
number goes at the top) rather than graph style (i.e. lowest y value
goes at the bottom) -ysc(reverse)- does what you want.

Not every graph needs a different name, and if there isn't a name for
this that's a positive too.
As different graphs can have the same name, and vice versa,
terminology can just make trouble for everyone. I've seen the term
"line plot" which I thought was always just segmented lines used for a
scatter of point symbols along one line!

Here's some code to get you started on your graph.

twoway rspike sympstart lastfu id, horizontal || ///
scatter id admdate || ///
scatter id firstpcr || ///
scatter id lastpcr  , ysc(reverse) yla(1/9, ang(h) noticks)  ///
legend(order(2 "admdate" 3 "firstpcr" 4 "lastpcr") pos(3) col(1))

Simple tip, perhaps superfluous: Put graph commands this complicated
in the do-file editor  giving yourself plenty of space and run from
there. Save recipes you want to repeat as self-contained do files that
would start by reading the data in.

Nick

On Fri, Mar 4, 2011 at 10:33 AM, Jannik Helweg-Larsen
<jhelweg@dadlnet.dk> wrote:

> I'm searching for a way to illustrate  horizontal timelines with
> markers for individual  time events, preferably in line with this very
> nice graph- but been unable to identify the right kind of graph
>
> http://cid.oxfordjournals.org/content/52/suppl_1/S102/F1.expansion.html
>
> Although I realise that this kind of  detailed graphical display is
> not routinely available- I would like to find som sort of graphing
> which comes close to this example:
>
> Is there a name for such a graph?
>
> To give a concrete example (real data):
>
> For each individual influenza patients hospitalized I would like to
> graph time (days) on horizontal lines starting from onset of symptoms
> illustrating  the total time of hospitalisation, with indicators along
> these lines with:
>
> 1) Admission days after symptomstart
> 2) First day PCR positive for Influenza
> 3) Last Day PCR positive
> 4) Days after symptomstart  starting tamiflu
> 5) Days after symptomstart  stopping tamiflu
>
>  id   sympstart     admdate    firstpcr     lastpcr   tamistart
> tamistop      lastfu
>  1.   10/12/2010  12/12/2010  22/12/2010  04/01/2011  22/12/2010
> 10/01/2011  21/01/2011
>  2.   18/12/2010  21/12/2010  31/12/2010  03/01/2011  02/01/2011
> 08/01/2011  10/01/2011
>  3.    01/01/2011  06/01/2011  06/01/2011  17/01/2011  07/01/2011
> 12/01/2011  27/01/2011
>  4.   03/01/2011  13/01/2011  13/01/2011  17/01/2011  14/01/2011
> 18/01/2011  25/01/2011
>  5.   04/01/2011  06/01/2011  16/01/2011  19/01/2011  17/01/2011
> 19/01/2011  19/01/2011
>  6.   16/01/2011  17/01/2011  20/01/2011  27/01/2011  23/01/2011
> 28/01/2011  14/02/2011
>  7.   23/01/2011  24/01/2011  24/01/2011  03/02/2011  24/01/2011
> 07/02/2011  21/02/2011
>  8.   28/01/2011  04/02/2011  06/02/2011  11/02/2011  07/02/2011
> 15/02/2011  26/02/2011
>  9.   28/01/2011  01/02/2011  02/02/2011  10/02/2011  01/02/2011
> 14/02/2011  01/03/2011

*
*   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/