https://issues.apache.org/ooo/show_bug.cgi?id=124075
Regina Henschel <rb.hensc...@t-online.de> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|UNCONFIRMED |RESOLVED
CC| |rb.hensc...@t-online.de
Resolution|--- |NOT_AN_ISSUE
--- Comment #2 from Regina Henschel <rb.hensc...@t-online.de> ---
The x-axis is formated as days. Those are used with there internal number for
purpose of regression calculation. Therefore the regression calculation is
based on the table
Date St. Joseph, Mo.
41629 782
41645 965
41660 948
41676 1181
41691 1414
41705 1633
41719 1852
A power regression is a curve of type
y = b * x^a
The parameters a and b are not calculated directly but the equation is
transformed to
ln(y) = ln(b)+a * ln(x)
Then from this linear model the regression is calculated. This gives
ln(a) = 399,4854456208
ln(b) = -4242,4976650638
You can proof this values by applying ln to your data and calculating the
linear regression parameters using LINEST.
The parameters a and b are calculated from this results by applying the inverse
function.
a = exp(399,4854456208) = 3,1212E+173
b = exp(-4242,4976650638) = 0
b is so near to zero, that it is not possible to show the difference in the
number format 'double', which is used in Calc, and therefore the result is
zero.
When you now insert the values of a and b in the first equation you get
y = 0 * x^3,1212E+173
and that simplifies to
y = 0.
Older versions hide this problems for line charts, because line charts were not
able to tread date values as number, but used them as pure (text) categories.
For line charts a dummy x-value was used, counting the category values 1, 2, 3,
4, ...
To get proper regression equations you have to change the dates to a numerical
value "days since <zero day>". When you expect a power regression for some
theoretical reasons, you should set <zero day> not too far in the past to get
values in the range, which can be expressed by 'double' number format, which is
approximately +- 10^+-308.
So this result in the chart is no error.
--
You are receiving this mail because:
You are on the CC list for the bug.
You are the assignee for the bug.
You are watching all bug changes.
