Re: [O] Generating boxplot (candlestick) diagrams (babel or org-plot)

[Eric Schulte]

Karl Voit <address@hidden> writes:

> Hi Eric!
>
> * Eric Schulte <address@hidden> wrote:
>>
>> In more complex examples like these I would recommend using a gnuplot
>> code block rather than using Org-plot.  
>
> Thought so. Thanks for confirmation.
>
>> Gnuplot has an extremely
>> comprehensive built in help system, typing "help boxplot" at the gnuplot
>> REPL returns the following.
>
> My gnuplot (Version 4.4 patchlevel 2) returns:
>
> ,----
> | gnuplot> help boxplot
> | Sorry, no help for 'boxplot'
> | gnuplot>
> `----
>
> gnuplot-doc is installed. Do I miss something else? «help
> candlestick» works. Is «boxplot» a part of an additional extension
> package?
>

I'm using gnuplot 4.5 patchlevel 0 which is probably where this
difference comes in.  As I use gnuplot all the time, I find it
worthwhile to track the development head for nice new features such as
this and beautiful pdf plots through pgf/tikz.

>
>> Also see the help for "candlestick".  Using this information the
>> attached Org-mode file generates a candlestick plot using a Gnuplot code
>> block.
>>
>> #+name: sample
>>| x | box_min | whisker_min | whisker_high | box_high |
>>|---+---------+-------------+--------------+----------|
>>
>> #+begin_src gnuplot :var data=sample :results silent
>>   set xrange [0:7]
>>   plot data using 1:2:3:4:5 with candlesticks
>> #+end_src
>
> Your example works like charm at my side. 
>
> And now I also understand the problem I had: I thought that gnuplot
> is able to derive min/max whisker min/max from the data columns by
> itself.
>
> Well - this is bad news to me since I do not see any reasonable way
> to use gnuplot for that purpose. I might need to learn R or
> Python/matplotlib :-(
>

Judging from the help output for boxplot it looks as though the boxplot
option *will* perform these calculations for you, if you are willing to
move to version 4.5 or later.  The only downside of using new features
in gnuplot is that it is noticeably harder to find help through google.

>
> Thanks for your help anyway!

Happy to help!

-- 
Eric Schulte
http://cs.unm.edu/~eschulte/