group fit

group fit

Post by crawford richa » Fri, 23 Jan 1998 04:00:00





> I've got a few data files  (df1,df2,...) and corresponding fixed
> parameters a1,a2,...
> I also have a function, e.g. f(x,a)=a+u+v+.. and some free parameters
> u,v,...

> Fitting and plotting with only one set of data isn't a problem:

>         fit f(x,a1), 'df1' via u,v,,...
>         plot f(x,a1), 'df1'

> But I need to fit u,v,... with the whole data (don't forget a1,a2,..)
> Furthermore I have to plot all data with the referring fitted curve.

> If I am right that is not the same problem Loris Benett had or has. But
> it is more general.

> Is it possible to solve my problem with gnuplot? And how?

gnuplot has a concept of a multi-data-set file, but it has no concept
of a multi-file data set -- all of the data to be treated as a single
data set must be in the same file.

You don't necessarily need to save the all-data file ... if your system
supports redirection, you might be able to do everything from within
gnuplot.  In Unix, for example, if you want to fit all the data in files
named "datafile.nnn", you could use
    fit f(x,a1) '< cat datafile.*'

If your files aren't named in a convenient way or you need to pull different
columns from different files, you could write a little program that extracts
the data and writes it to STDOUT and then run it from within the datfile
field in the same way.

I don't know if this helps ...


 
 
 

group fit

Post by Hans-Bernhard Broek » Sat, 24 Jan 1998 04:00:00


[Peter: ich antworte in Englisch, weil's auch an die Newsgroup geht]


> I've got a few data files  (df1,df2,...) and corresponding fixed
> parameters a1,a2,...
> I also have a function, e.g. f(x,a)=a+u+v+.. and some free parameters
> u,v,...
> Fitting and plotting with only one set of data isn't a problem:
>         fit f(x,a1), 'df1' via u,v,,...
>         plot f(x,a1), 'df1'

OK, so I assume it's gnuplot pre-3.6 we're talking about (gnuplot-3.5
doesn't have 'fit').

Quote:> But I need to fit u,v,... with the whole data (don't forget a1,a2,..)
> Furthermore I have to plot all data with the referring fitted curve.

Your problem has to aspects. First of all, you'll have to transform
your multiple datafiles into *one*. Second, you'll have to do a
multi-branch fit, or equivalently, a 2-D fit.

The combined datafile will have to be in 2-D format, i.e. each time
a new df{i} starts, separate it from the previous one with one blank
line. It'd also be advantageous to have the a{i} values corresponding
to each df{i} as an additional column in the combined datafile.

Once all that is done, you can do the fit this way:

    fit f(x,y) 'bigdatafile' using 1:2:3:4 via u, v

(or similar). If you have written an automatic program that combines
the datafiles and writes the result to stdout, you may write
'<theprogram' instead of 'bigdatafile'

Please also see the example case for this kind of fit in the gnuplot

--

Even if all the snow were burnt, ashes would remain.

 
 
 

1. Group Fit ( where described )

Hi All,

I am a statistical newbie but since my business is systems performance I
am taking the risky step of being smarter than my boss and applying
formulae, often translated to perl, to my collected data.

I would be most grateful if you could point me either to 3.6 stuff or
other sources that might help.  So far I am relying on exceptions from
threshold triggers, using linear regression and expn smooting for
forecasting.  These have their problems though, because systems are
fraught with outliers, and these are actully important.

The hugeness of this type of data and streaming batch requirements (a
contradiction) preclude any $1K+ packages entering the picture.  Every
thing has to be written in perl.

Any clues would by most helpful, thanks in advance.

Yours, John van V.

______________________________________________________
Get Your Private, Free Email at http://www.hotmail.com

2. Clipping a polygon...

3. Beizer curve fitting/ B-spline Fitting

4. Parsing/converting SL with lex/yacc

5. Gimp FITS violates FITS standard.

6. TImage

7. Gimp FITS images violate FITS standard

8. .swf flash files seen under a TImage and TOpenPictureDialog

9. COVERSION FROM TIFF GROUP 3 TO TIFF GROUP 4

10. NEW GROUP - Forming New Group on Scientific Visualization

11. Call For Group: comp.graphics.3d (Was: PoV Ray Related Group NEEDED)

12. seeking info on viewing CCITT group 3 / group 4 tiff images

13. Fax group 4 and group 3 compression