Bug#741626: qtiplot: Fits processed under Fit Wizard give wrong results

[Burkhard Bunk]

Package: qtiplot
Version: 0.9.8.8-5+b1
Severity: important

Dear Maintainer,

starting from Wheezy, qtiplot gives wrong fit results when using the
Fit Wizard.

Here is my dataset (which I use to test qtiplot since 2007):

    x_i    y_i    s_i		s_i: error of y_i (standard deviation)
    ------------------
    0.68   3.92   0.41
    1.81   6.80   0.67
    2.57  10.24   0.80
    3.41  12.15   0.92
    4.85  16.02   1.10
    5.76  19.94   1.20
    6.90  21.83   1.31
    7.84  26.61   1.39
   -------------------

Go for a (linear) maximum likelihood fit

    y = a*x + b

The correct results, following well-known rules, are

    a = 3.055892 +/- 0.129601
    b = 1.770606 +/- 0.410934

    chi^2 / doF = 2.694952 / 6.000000 = 0.449159

These results were produced by qtiplot as well (always using
"instumental weighting") -- up to Squeeze (qtiplot-0.9.8-1).

In Wheezy (qtiplot-0.9.8.8-5+b1), using the Fit Wizard, I obtain

    a = 3.1003594093891e+00 +/- 1.8592231846202e-01
    b = 1.5497365804245e+00 +/- 7.7928406686591e-01
    Chi^2/doF = 4.3061339571571e-01

which is pretty much off the correct values.

This is reproducably so, whenever the Fit Wizard is used, no matter
whether the "Built-in" linear function is chosen or a "User defined"
function of the same form (and "Instrumental Weighting" in all cases).

It should be noted that the correct result is still obtained with the
"Fit Linear" option selected directly in then "Analysis" menu (without 
recourse the the Fit Wizard).

Exactly the same thing happens in Ubuntu-12.04 (which has the same qtiplot
version as Wheezy).

The Fit Wizard is crucial, however, for fit functions beyond the set
of Built-ins and for a precise control of the fit procedure, especially in
the non-linear case.
And a program for scientific data analysis should not produce wrong results,
of course.
Hence the severity of "important".

It looks like a degression within the Fit Wizard, not the fit code itself.
Any idea or workaround?

A fix would be highly appreciated.
If you need more info or help in testing, please let me know.


-- System Information:
Debian Release: 7.4
   APT prefers stable-updates
   APT policy: (500, 'stable-updates'), (500, 'stable')
Architecture: amd64 (x86_64)

Kernel: Linux 3.6.11-X86_64 (SMP w/4 CPU cores)
Locale: LANG=C, LC_CTYPE=en_US.utf8 (charmap=UTF-8)
Shell: /bin/sh linked to /bin/bash

Versions of packages qtiplot depends on:
ii  libalglib-2.6.0           2.6.0-6
ii  libc6                     2.13-38+deb7u1
ii  libgcc1                   1:4.7.2-5
ii  libgl1-mesa-glx [libgl1]  8.0.5-4+deb7u2
ii  libgl2ps0                 1.3.6-1
ii  libglu1-mesa [libglu1]    8.0.5-4+deb7u2
ii  libgsl0ldbl               1.15+dfsg.2-2
ii  libmuparser2              2.1.0-3
ii  liborigin2-1              2:20110117-1+b2
ii  libpng12-0                1.2.49-1
ii  libpython2.7              2.7.3-6
ii  libqt4-network            4:4.8.2+dfsg-11
ii  libqt4-opengl             4:4.8.2+dfsg-11
ii  libqt4-qt3support         4:4.8.2+dfsg-11
ii  libqt4-svg                4:4.8.2+dfsg-11
ii  libqt4-xml                4:4.8.2+dfsg-11
ii  libqtassistantclient4     4.6.3-4
ii  libqtcore4                4:4.8.2+dfsg-11
ii  libqtexengine1            0.3-3
ii  libqtgui4                 4:4.8.2+dfsg-11
ii  libstdc++6                4.7.2-5
ii  libtamuanova-0.2          0.2-2
ii  python                    2.7.3-4+deb7u1
ii  python-qt4                4.9.3-4
ii  python-sip                4.13.3-2
ii  zlib1g                    1:1.2.7.dfsg-13

Versions of packages qtiplot recommends:
pn  qt-assistant-compat  <none>
ii  qtiplot-doc          0.9.8.8-5

qtiplot suggests no packages.

-- no debconf information

Regards,
Burkhard Bunk.
----------------------------------------------------------------------
  bunk at physik.hu-berlin.de      Physics Institute, Humboldt University
  fax:    ++49-30 2093 7628     Newtonstr. 15
  phone:  ++49-30 2093 7980     12489 Berlin, Germany
----------------------------------------------------------------------