Hi to everyone,
I have greatly expanded the wiki page:
- there are a number of new Chart types
-- both higher priority (Histograms, Venn-diagrams, and others
previously described)
-- as well as a huge number of lesser priority charts
- I have described some NEW *High*-*Priority* Issues
-- Significance Values (p-values)
-- Outliers [aka splitting the Y-axis]
Please read first the wiki.
*p-Values*
=======
One of the *most useful* informations present on a Chart which is way to
*often forgotten* describes the significance of the displayed difference.
- a trivial difference (i.e. non-significant) might look frightening big
through mischoosing the axes
- while a real difference might be overlooked
This feature conveys really additional information (unlike some other
techniques, like voluminizing)!!!
*Outliers*
=====
Occasionally we have to represent data that contains outliers. This is
not easily performed and distorts easily the data. I have expanded on
the wiki page a possible solution (allow Y-Axis to be broken into
independent sections, e.g. 0-20 and then 200-220, so that values of 0-10
can easily be displayed in real proportion, while an outlier of 200
won't distort the whole chart).
IMPRESS
=======
Is this mailing list used to discuss Impress feature, too?
I have some more global comments and suggestions for Impress, too, that
are not easily covered with Bugzilla (and which would have great impact
on many aspects of Impress).
--------------
Some comments on previous posts:
BTW: THERE IS NO REASON TO *SHOUT* HERE !!!!
I hope not to be misunderstood. It was and is NOT my intention to shout.
It is my intention to highlight very useful keywords.
This is a proven method. Some 90% of students highlight specific words
when learning (using some marker or pen). I did it myself when I was a
student, and now I am using it even more often when I read a scientific
article (and in my life I have read thousands of scientific articles; I
have more than 3000 articles on my PC). You do not have always time to
read everything accurately, and knowing which words are particularly
important is of great help.
You cannot be earnest, that a warning dialog pops up every time a user
choses a 3d type.
Well, my browser (SeaMonkey) also pops a warning dialog when I browse a
secure site. It is possible to disable the warning (check the "Don't
show me this anymore" box), so why should it be different in Chart.
People are unaware of scientific statistics and charting, so a little
more education for the users will in the end only make OOo more popular.
... the chef said: "Oh I wouldn't recommend that, take something else".
The chef didn't learn marketing. He should have said: "Our specialty is
... (this). I would strongly recommend this one ... and so on, surely
NOT to say, don't take this. How does this apply here: of course the
warning should be drafted in such a way, that the user understands that
3D is wrong and 2D offers great advantages, NOT that just don't take 3D.
e.g., one advantage of plain charts:
- it is more user friendly for the audience, see the example below:
- try to read the exact value on a 3D bar: Where do you read it?
- you need to read at the back of the bar, go from there horizontally
to the Y-axis;
- on the intersection with the posterior portion of the Y-axis draw a
slanted line that goes to the front of the Y-axis (the legend is
actually on the front of the 3D Y-axis; at what angle to you draw the
slanted line?)
- so, how fast can you actually read the Y-value? Was it accurate?
[Most bars I saw did NOT explicitly display magnitudes, so one must read
them from the Y-axis.]
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.
I will try to write such a hint, however, please note, that while I know
much statistics (and I even teach it), I am a physician (my specialty
are infectious diseases and tropical medicine), so a renowned
statistician is surely more able to do the work. Unfortunately, Prof.
Alvan Feinstein died in 2001
(http://www.yale.edu/opa/v30.n9/story12.html). He was a strong critic of
many lamentable statistical approaches.
- very small values [proportions] are NOT easily displayed and NOT
easy to see [in pie charts]
Then don't use them.
I described there a general limitation of pie-charts. There are some
wokarounds (see points 6.h-6.i) dealing with this. There are other
workarounds. Try to display a proportion of "0" in a pie chart; it
simply won't work. Yet pie-charts are the most used chart for
proportions. So a brainstorming session to expand the pie-chart concept
for this situation is somehow needed.
Another problem are charts with negative values. ...
This doesn't sound like a very common thing to do.
While I have described a very particular situation (values added up to
0), the real problem is with *negative proportions*. It does not make
sense to mix them with positive proportions. SO, group positives with
positives, and negatives with negatives. (<-- for pie charts; for
bars/percent chart see also point 6.e)
Don't get me wrong: Your suggestions for scientific charts are very
welcome, just please don't demand removing features that are important
to other users.
I hope I didn't got wrong. It is NOT about removing, it is about setting
priorities.
At the risk of stating the obvious: There are people in the world to
whom "marketing" is daily work instead of an insult, and their
requirements have to be taken into account, too.
I have seen wonderful bar-charts and pie-charts that were 2D (and none
that was 3D), so it is possible to get majestic results even with 2D
charts (and actually these results are even better). The problem is *the
default* styles used in Chart are awful. My recommendation: implement
styles in Chart and develop better styles (especially colours, including
different colour sets for charts with only 2 data-sets, 3 data-sets, 4
data-sets, i.e. use custom colour-sets for a specific number of colours
- DO NOT lump everything together; ...)
[the worse pie-chart example]
That quote refers to a type of chart we don't have, where the values
are represented by the height (thickness) of the segments. Consider it
a bar chart in circular layout and it even looks kind of neat.
While the quote was specific for one particular type of pie-chart, its
sense is more broadly and the worst-chart example is just the upper [or
lower] limit of bad pie charts. Volumizing is always wrong, when the
same data can equally well be presented using a 2D chart. If one reads
carefully Chapter 9 from the book (and actually the whole book), this
becomes more clearly.
Sincerely,
Leonard Mada
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
