0README

Wed Jul 27 22:42:31 MST 2005


|-- Line			directory - sources, executable, tests:
    |					    function for linear least-sq fit
    |					    of test set of data to straight line
    |					    compare with hand calcutions
    |-- 0NOTE
    |-- 0README
    |
    |-- Makefile
    |-- linefit				executable - function + simplex library
    |-- linefit.c			C-code source for function
    |
    |-- linefit.dat			data, parameters - input for linefit
    |
    |-- linefit.out			results of fit
    |
    `-- tests				directory - various tests of linefit
        |-- linefit.test1-4d30
        |-- linefit.test1-mod-4d30		1991 results, under 4d30,
        |					    edited as above
        |-- linefit.test2-4d30
        |-- linefit.test2-mod-4d30
        |-- linefit.test2003-octane
 	|
        |-- test.data
 	|
        |-- test.out
        `-- test.out-macosx			current results, under macosx



Linefit permforms a linear least-squares fit of a manufactured data set to a
straight line.


Compare simplex code with hand calculation.

	vi linefit.out linefit.dat

		intercept and slope are exactly same, cf hand and simpfit
		standard deviations of parms diff by 1-2% (3 sig. digit)



Compare Mac OS X with sgi compiled code.
----------------------------------------
linefit under sgi octane:
mse    = 6.44040404040405e-02   rms weighted error = 2.53779511395307e-01

yzero  = 5.79636363636365e-01
ymin   = 5.79636363636365e-01
y-pmin = 5.79636363636364e-01

parm free    q                 centroid          pmin              std_dev
  0   0  1.4844472718e-07  1.0109090909e+00  1.0109090915e+00  1.6244739204e-01
  1   1  2.6143622822e-08  1.9981818182e+00  1.9981818181e+00  2.3877623977e-02
----------------------------------------
linefit under macosx:
mse    = 6.44040404040406e-02   rms weighted error = 2.53779511395307e-01

yzero  = 5.79636363636366e-01
ymin   = 5.79636363636366e-01
y-pmin = 5.79636363636364e-01

parm free    q                 centroid          pmin              std_dev
  0   0  1.4844472710e-07  1.0109090909e+00  1.0109090900e+00  1.6169682647e-01
  1   1  2.6143622674e-08  1.9981818182e+00  1.9981818183e+00  2.3894826536e-02
----------------------------------------

	rms weighted error		identical
	centroid final simplex parms	identical
	quad fit minimum parms		differ 9th sig digit
	standard deviations		differ 4th sig digit
	function val quad fit min	differ 15th sig digit



John Rupley
 rupley@u.arizona.edu -or- jar@rupley.com
 30 Calle Belleza, Tucson AZ 85716 - (520) 325-4533; fax - (520) 325-4991
 Dept. Biochemistry & Molecular Biophysics, Univ. Arizona, Tucson AZ 85721
