[Libreoffice-bugs] [Bug 142643] LibreOffice Calc power regression coefficients error

bugzilla-daemon at bugs.documentfoundation.org bugzilla-daemon at bugs.documentfoundation.org
Sat Jun 5 10:28:48 UTC 2021


https://bugs.documentfoundation.org/show_bug.cgi?id=142643

--- Comment #15 from Uwe Auer <uwe at franke-auer.de> ---
Created attachment 172638
  --> https://bugs.documentfoundation.org/attachment.cgi?id=172638&action=edit
Power Regresssion Methods - Comparision of Coefficients

I'm not sure whether this is really a pure documentation issue only, but indeed
a problem of the "Data -> Statistics -> Regression -> Type: Power Regression"
functionality (at least from a users perspective). 

My comments are based on how I do such power regression, which essentially is
based on using:

[1] =SLOPE(LN(range_of_Y_values);LN(range_of_X_values))
[2] =EXP(INTERCEPT(LN(range_of_Y_values);LN(range_of_X_values)))

according to function  LN(f(x)) = LN(a) + b * LN(x) (derived from f(x)=ax^b)

As a user I'd expect a "Power" regression analysis to yield the coefficients a
and b directly (i.e. without any further exponentiation) just like a power
trend line in a chart shows both coefficients correctly (without applying any
further math) setting option "[x] Show Equation". So it least when comparing 

[1] Coefficients in trend line equation in chart --with--
[2] Coefficients in "Data -> Statistics -> Regression -> Type: Power
Regression" (see cells J20 and J21 in the sample file) 

there is a discrepancy. See also attachment.

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/libreoffice-bugs/attachments/20210605/8ecf7c25/attachment.htm>


More information about the Libreoffice-bugs mailing list