Hi Leonard,
First thanks a lot for checking the alpha version of the chart
reimplementation!
Leonard Mada wrote:
Hello
What in particular disturbed you about the 3d charts in the new chart
(That's what I suppose you mean by new 3D pie-charts)?
I was disturbed that so much effort is put into expanding these 3D
charts, when virtually any high-ranking statistician warns against their
use. To quote Prof. Feinstein: (applies to both 3D bars and 3D pies)
3D charts (also 3D pies) was an existing feature in the old chart
already. So it was no option to drop this whole feature area when
starting the reimplementation. Also the 3D charts were not 'expanded
that much' as you assume. What we did was a necessary consolidation of
the existing feature set. We needed to get rid of as much curiosities
and exceptions as possible to get a basis for further improvements. When
reimplementing a piece of software in most cases it does not save time
to try to make exactly the same errors that did exist in the former
implementation. The opposite is the case.
> A particularly lamentable custom is “volumizing”— a tactic that
converts the two-dimensional bar
> to a three-dimensional post. Since the second-dimension was needed
only for labeling, the third-dimension
> is completely unnecessary. Its use is an act of “marketing,” not
scientific communication.
> (Examples of these abuses for contrasts of groups will be shown later
in Chapter 16.)
And a somewhat similar concept: (the worse pie-chart example)
> The basic idea of a pie graph is abandoned because each category is
given about the same angular slice,
> and the circular shape has been converted to a quasi-ellipse for
portraying a third dimension.
Indeed, as I already stated, 3D charts DO NOT add any statistical
benefits and should be strongly discouraged!!!
Of course it does not add information, it's just that people like 3d
charts for presentation. Imagine what would happen if we suddenly
dropped the support for all 3d chart types. We can simply not do this.
And besides, we don't want to.
I believe it is the duty of a good program to *warn* the user when he
takes the wrong decision. Let the 3D charts sit there, BUT when a user
selects one for use, warn him about this and offer an explanation, why a
third dimension is not desirable. I would also suggest to ask a number
of high-ranking statisticians to wright this explanation, so that - at
least some of the users - would abandon the 3D chart use.
Hm, I tend to assume that a user should be able to decide himself
whether a 3D pie is appropriate for his data or not, so I don't want to
bother him with repeated warnings that he might do something stupid. But
maybe there could be placed a hint into the help for charts. What do you
think? Would you mind to write an according issue with a suggestion for
the warning text? And send it to iha. Thanks a lot.
e.g. Tell me please, which is bigger, 'Column B' or 'Column C':
You will wonder, that both segments are equal (38%). I created this pie
chart with the *default* settings in Calc, so it is really NOT me
choosing a particular very wrong setting. IT WILL ALWAYS HAPPEN when
having two similar values that together add to more than 50% (=> one
segment will be drawn in the back and will look rather small, and the
next one comes in the front, looking rather big).
Another great problem arises when one has to compare segments from *2*
*different* *pies*:
- IF the segments are not located in the same position (e.g. 3 o'clock
on one pie, BUT 6 o'clock in the 2nd one), you would again come into big
trouble.
Of course, you should always use 3d charts with care. If you have a 3d
pie which you view almost from above, there is no big difference to a
2d pie, except that <<some people find this more appealing.>>
<<Some people>> means most (if not all) significant statisticians, that
is the difference.
PROBLEMS
=========
There are a number of problems even with the standard pie-chart (2D). I
mentioned one in my first post.
- very small values are NOT easily displayed and NOT easy to see
See Issue 64695: Pie chart with a second pie chart
http://www.openoffice.org/issues/show_bug.cgi?id=64695
and Issue 64694: Pie chart with a second bar chart
http://www.openoffice.org/issues/show_bug.cgi?id=64694
Would you rate this higher than Regression Curve parameter display for
example?
Therefore I would recommend putting some effort into solving this problem.
I would also like to be able to rotate (clockwise or counter-clockwise
in 2D, NOT 3D) the pie chart (this is especially useful, when one wants
to align 2 pie charts).
See Issue 37823: pie chart in Calc draws segment in opposite direction
with Excel
http://www.openoffice.org/issues/show_bug.cgi?id=37823
Also, I would sometimes want to swap 2 segments.
See Issue 64522: reordering the segments in a pie chart
http://www.openoffice.org/issues/show_bug.cgi?id=64522
Another problem are charts with negative values. I wanted recently to
create a pie chart, highlighting the difference between 2 groups (2 pie
charts), so I created the difference between the proportions. SOme were
positive, some were negative. I would have welcomed a mechanism to
automatically group the positive proportions vs the negative proportions
(they must sum up to 0, so both groups would have occupied half of the
piechart). Instead, the values were used as absolute values AND
therefore were intermixed, bringing NO meaning.
Would you mind to write an issue with a suggested behaviour? Thanks a lot.
There are many other issues that could be improved.
That's true. In a first step we will finish the reimplementation with a
feature set as is. Then we can improve further. So when you find errors
in the chart reimplementation that did not exist in the old chart please
share them! Thanks a lot.
Kind regards,
Leonard Mada
Kind regards,
Ingrid
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
