[sci.bio] GAUSSIAN FITTING PROGRAM

appel@cui.unige.ch (APPEL Ron) (02/07/91)

I am looking for a public domain program, doing the following:

Given: a discrete representation of a hill shaped 2D curve,
find:  the best fitting gaussian curve.

Thanks for giving me some pointers as to where I can find that program.

####################################################################
Ron D. Appel                              e-mail: appel@cih.hcuge.ch
Unite d'Imagerie Numerique                  
Centre d'Informatique Hospitaliere
Hopital Cantonal Universitaire de Geneve                    
24, rue Micheli-du-Crest                                 ###
CH-1211 Geneve 4                                     ########
Switzerland                                      ###############
tel.: 41-22-229254                            ###################
fax : 41-22-227073                      ##########################
####################################################################

toms@fcs260c2.ncifcrf.gov (Tom Schneider) (02/12/91)

Sorry if this posting occurs twice.  Something was odd about the listed
newsgroups and I think it didn't go out the first time.

In article <4907@cui.unige.ch> appel@cih.hcuge.ch (APPEL Ron) writes:
>I am looking for a public domain program, doing the following:
>
>Given: a discrete representation of a hill shaped 2D curve,
>find:  the best fitting gaussian curve.
>
>Thanks for giving me some pointers as to where I can find that program.

One solution that might help is to calculate the mean and standard deviation of
your distribution.  This immediately defines a gaussian which will fit your
data.  I don't know whether or not it is the best fitting curve (Hey Jeff
Haemer:  what do you say?? :-), but it obviously will fit the data very well if
your data DO have a gaussian distribution!  You can determine if your fit is
good by a chi-square test between a histogram of your data and the
gaussian from the mean and standard deviation.

I have placed a program called genhis in an archive for you.  You can obtain it
by anonymous ftp from ncifcrf.gov in directory pub/delila.  The program is in
Pascal: genhis.p, and has an example parameter file: genhisp.  (Both files are
compressed, so are called genhis.p.Z and genhisp.Z in the archive.)  The
program probably can be automatically translated to C if you do not have a
Pascal compiler.  You give genhis raw numbers and it generates a histogram,
calculates mean, standard deviation and makes a graph for you.  It was written
by Gary Stormo.  Documentation is in comments in the program code itself.  I
use this tool all the time; it is extremely useful!

>####################################################################
>Ron D. Appel                              e-mail: appel@cih.hcuge.ch
>Unite d'Imagerie Numerique                  
>Centre d'Informatique Hospitaliere
>Hopital Cantonal Universitaire de Geneve                    
>24, rue Micheli-du-Crest                                 ###
>CH-1211 Geneve 4                                     ########
>Switzerland                                      ###############
>tel.: 41-22-229254                            ###################
>fax : 41-22-227073                      ##########################
>####################################################################

  Tom Schneider
  National Cancer Institute
  Laboratory of Mathematical Biology
  Frederick, Maryland  21702-1201
  toms@ncifcrf.gov