Re: new rfunctions

[<pl...@pi...>]

On Wed, 23 Jan 2008 02:57:31 +0100, Philipp K. Janert <ja...@ie...>
wrote:

>
> I am also not sure that having min(), max(),
> etc for some fixed number of args would be
> all that useful.
>
> What would be AWESOME, however (at least
> in my opinion) is to have those kinds of functions
> to operate on input data.
>

Well that is getting into the realm of data processing that some feel is=

beyond the scope of a plotting tool and indeed the line has to be drawn
somewhere.

This sort of thing can already be done using the new assignment syntax b=
ut
requires two passes. ie a first plot command that calls some function on=

each point as it is plotted , then a new plot command that uses some
stored result and plots what you want to see.


> Why? Here is an example. Often I would like to
> do this:
>
> plot "data" u 1:($2 - mean($2))
>
> or
>
> plot "data" u 1:( ($2-mean($2))/stddev($2) )
>
> etc.
>

mean_sd(x)=3D (....... , mean=3D..., sdddev=3D...);

   plot "data" u 1:(mean_sd($2))
   plot "data" u 1:( ($2-mean)/stddev )

I currently do something like this to plot an arrow representing the mea=
n
y value. The arrow has to be defined before the plot so I have to do one=

run just to get the mean with a trivial function then replot with my arr=
ow.

> Given how data file input is currently handled,
> something like the above would be difficult to
> implement, I think (would require multiple passes).
>
> A very useful alternative might be to have a
> "stat" command, so that I could say:
>
> stat "data" u 1
>
> and it prints a brief report to the terminal, including
> all the usual suspects (mean, stddev, min, max, median...)
> That would be tremendously useful when working with
> data...
>

That would be an never ending list of desired functions that someone wou=
ld
always want extending.
There exists now two mechanisms , one internal and of course the
traditional ablility to call any outside prog or script to do more compl=
ex
analysis.

Maybe someone with a good grounding in these stats could provide a set o=
f
gnuplot functions as a .gnu that could be added to any users .gnu with o=
ne
line of load "basic-stats.gnu" like Ethan suggested.

If something is to be added here like your stat suggestion it could be a=

dont_plot option that mimics the way plot goes through the data without
spending time creating the output. This would allow a prepass for simple=

d.p. functions as outlines above without much of the redundancy.

I expect it would be pretty simple to add this as an plot option or null=

line type that simple avoids calling the output routine. I'd have to dig=

into the code.

The above example could become:

   plot "data" u 1:(mean_sd($2)) noplot
   plot "data" u 1:( ($2-mean)/stddev )

This could substancially improve the efficiency of the dual pass appraoc=
h.


regards, Peter.


> Best,
>
> 		Ph.
>
>
> On Tuesday 22 January 2008 14:41, Ethan Merritt wrote:
>> On Tuesday 22 January 2008 14:30, Mojca Miklavec wrote:
>> > On Jan 22, 2008 11:20 PM, Ethan Merritt wrote:
>> > > On Tuesday 22 January 2008 05:54, Daniel Heiserer wrote:
>> > > > I would like to add some functions to the CVS version of gnuplo=
t.
>> > > > These functions include:
>> > > > min()
>> > > > max()
>> > > > std()
>> > > > mean()
>> > > > etc.
>> > >
>> > > Could you explain why you need to have these as built-in function=
s?
>> > > What is wrong with just defining them yourself:
>> > >
>> > >   min(A,B) =3D A < B ? A : B
>> > >
>> > > You can put that definition in your customization file ~/.gnuplot=

>> > > if you always like to have it.
>> >
>> > I agree that min, max, mean and others might be handy to have. I
>> > always "hack them" with ($1+$2+$3+$4)/4 or the way you have just
>> > shown, but these are really basic functions that are often used and=

>> > easy to implement.
>>
>> The two argument case is trivial, as shown above.
>> The general case of N arguments would be less trivial, particularly i=
f
>> you are looking for a way to make the calculation over an arbitrary
>> number of data points read from a data file.
>>
>> Before jumping in with both feet, I think it would be useful to
>> block out what exactly is the goal.
>>
>> I can understand wanting MIN and MAX for simple bookkeeping,
>> for example choosing which of two column values to use.
>>
>> But the need for a command line  std(a,b,c,...) or mean(a,b,c,...)
>> function is not obvious to me.  This seems more like something one
>> would track automatically during the course of data input than
>> something one would type in manually at the command line.
>>
>> > Instead of asking "why one would want to add them", I would rather =
ask
>> > "why not".
>>
>> The reason to ask "why" is that there may be a better way to achieve
>> the actual goal.
>
> ----------------------------------------------------------------------=
---
> This SF.net email is sponsored by: Microsoft
> Defy all challenges. Microsoft(R) Visual Studio 2008.
> http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
> _______________________________________________
> gnuplot-beta mailing list
> gnu...@li...
> https://lists.sourceforge.net/lists/listinfo/gnuplot-beta
>