Bug 46539 - sigmoid curve fitting in trend lines
Summary: sigmoid curve fitting in trend lines
Status: NEW
Alias: None
Product: LibreOffice
Classification: Unclassified
Component: Chart (show other bugs)
Version: unspecified
Hardware: All All
: high enhancement
Assignee: Not Assigned
QA Contact:
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2012-02-23 14:13 UTC by James Salsman
Modified: 2012-02-23 18:21 UTC (History)
1 user (show)

See Also:
i915 platform:
i915 features:


Attachments

Description James Salsman 2012-02-23 14:13:23 UTC
Exponentials do not actually exist in economics, biology, or any other finite field. They are all sigmoid curves such as the logistic or Gompertz curve. Offering users the ability to fit a trend line to an exponential curve but not a logistic and Gompertz sigmoid does them a disservice, and makes it much more likely that they will make extrapolation errors.

Here is how to do it in R: http://thebiobucket.blogspot.com/2011/04/fit-sigmoid-curve-with-confidence.html

and Matlab/Octave: http://www.mathworks.com/matlabcentral/newsreader/view_thread/169459

Thank you.
Comment 1 Regina Henschel 2012-02-23 14:35:53 UTC
The current standard ODF1.2 specifies only that kind of curves, which are implemented already. To get new kind of regression curves a addition to the standard is needed. Can you write a proposal to OASIS? It need not contain the algorithm but a reference to the mathematical definition is needed.

There have already been proposals for new kind of regression curves, so it is likely that additional types are possible for ODF1.3
Comment 2 James Salsman 2012-02-23 14:40:49 UTC
(In reply to comment #1)
> The current standard ODF1.2 specifies only that kind of curves, which are
> implemented already. To get new kind of regression curves a addition to the
> standard is needed. Can you write a proposal to OASIS? It need not contain the
> algorithm but a reference to the mathematical definition is needed.

Regina, I am very happy to do so, but please send me the URL to instructions if any, or at least the sub-organization involved, and email address(es) if you have them.

As for the mathematical definition, please see: http://en.wikipedia.org/wiki/Logistic_function

and http://en.wikipedia.org/wiki/Gompertz_function
Comment 3 James Salsman 2012-02-23 17:43:13 UTC
(In reply to comment #1)
> Can you write a proposal to OASIS?

Regina, I understand that I need to request a change to http://docs.oasis-open.org/office/v1.2/os/OpenDocument-v1.2-os-part1.html#__RefHeading__1419514_253892949 but where do I send that request?
Comment 4 James Salsman 2012-02-23 18:06:41 UTC
[done:]

Date: Thu, 23 Feb 2012 19:05:44 -0700
Subject: sigmoid curve fitting in trend lines
From: James Salsman <jsalsman@gmail.com>
To: robert_weir@us.ibm.com, office-comment@lists.oasis-open.org

Rob Weir
Chair
OASIS Open Document Format for Office Applications (OpenDocument)
Technical Committee

Dear Mr. Weir:

True exponential curves do not actually exist in economics, biology,
or any other finite field. They are all sigmoid curves such as the
logistic or Gompertz curve. Offering users the ability to fit a trend
line to an exponential curve but not a logistic and Gompertz sigmoid
does them a disservice, and makes it much more likely that they will
make extrapolation errors.

Therefore, I request that you modify OASIS Standard
OpenDocument-v1.2-part1, Section 20.43, "chart:regression-type" [
http://docs.oasis-open.org/office/v1.2/os/OpenDocument-v1.2-os-part1.html#_=
_RefHeading__1419514_253892949
] as follows:

At the end of the bulleted list under "The defined values for the
chart:regression-type attribute are:", please add:

"* logistic sigmoid =E2=80=93 Regression with a logistic sigmoid function =
=E2=80=93
approximate the values of the series using the model: y =3D
A+B/(1+e^(-(x-C)/D))

"* Gompertz sigmoid =E2=80=93 Regression with a logistic sigmoid function =
=E2=80=93
approximate the values of the series using the model: y =3D
A+B=C2=B7e^(C=C2=B7e^(D=C2=B7x))"

Please let me know by email to jps-oasis-regression@talknicer.com if
you have any questions about this request. Thank you.

Best regards,
James Salsman

Ref.: https://bugs.freedesktop.org/show_bug.cgi?id=3D46539
Comment 5 James Salsman 2012-02-23 18:10:08 UTC
Well, that looks terrible pasted into bugzilla.  The two lines in question are:


"* logistic sigmoid – Regression with a logistic sigmoid function –
approximate the values of the series using the model: y =
A+B/(1+e^(-(x-C)/D))

"* Gompertz sigmoid – Regression with a logistic sigmoid function –
approximate the values of the series using the model: y =
A+B·e^(C·e^(D·x))"
Comment 6 James Salsman 2012-02-23 18:21:37 UTC
(In reply to comment #5)
>...
> "* Gompertz sigmoid – Regression with a logistic sigmoid function –

s/logistic/Gompertz/ on that line. Sorry.


Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.