[Libreoffice-bugs] [Bug 133746] Conditional Colour Scale Formatting incorrectly rendered: MAX & MEAN colours are reversed

[bugzilla-daemon at bugs.documentfoundation.org]

https://bugs.documentfoundation.org/show_bug.cgi?id=133746

--- Comment #5 from Colin <that.man.colin at gmail.com> ---
(In reply to Mike Kaganski from comment #4)
> This is not a bug. Using Percentile for intermediate value, you ask it for
> such a value that would represent exactly half of the population see [1]
> that refers to [2]. Having total of 46 values in the data set used in the
> conditional format, and 26 of them being 1, there's no value that would
> represent half of the population (23 values). So it necessarily ends up with
> 1 as the percentile value, thus being equal to max, and so effectively being
> equal to 2-color scale, as the third color is never used.
> 
> Possibly what was wanted is using *Percent*, not *Percentile*.
> 
> [1] https://help.libreoffice.org/6.4/en-US/text/scalc/01/05120000.html
> [2] https://wiki.documentfoundation.org/Faq/Calc/142

I think I have to disagree. I have other spreadsheets that also have the same
scale of "fixed" percentage results 20,40,60,80,100 and they don't fail because
there is no 50%. It's not actually possible to divide a week by a whole number
of days to get 50% and I am only targeting the five weekdays which also imposes
the same impossibility. Also, the remedy as mentioned in my comment #3, whilst
setting the mean to "real" 0.5 never encounters a calculated 0.5 but gives a
"bug free" result.
Why does deleting any five random rows that also don't contain any 50% values
obviate the bug?
I'm not convinced the colour scale is a standard deviation function.

-- 
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/20200607/2caf07bd/attachment.htm>