Just wondered what people on here thought about this type graphic to display information about the number of events that have occurred over a 7-day period, broken down hour-by-hour?
[attachment=304:attachment]
Do you think it's an effect way of presenting this data, compared to say a conventional graph or bar-chart?
Any comments much appreciated.
Greg
Temporal analysis graphic
Started by
gregsd
, Feb 17 2006 03:29 PM
7 replies to this topic
#1
Posted 17 February 2006 - 03:29 PM
Greg Driver
GIS Analyst
MapInfo User...!
#2
Posted 17 February 2006 - 03:52 PM
I found this one especially easy to read (and was quite amazed by the results). A very interesting way of presenting the information, but it works for me.
Hans van der Maarel - Cartotalk Editor
Red Geographics
Email: hans@redgeographics.com / Twitter: @redgeographics
Red Geographics
Email: hans@redgeographics.com / Twitter: @redgeographics
#3
Posted 17 February 2006 - 04:18 PM
I like it - especially the clock analogy. I found it easy to read and to grasp patterns - much better than a bar chart could I think.
...my $0.02
e
...my $0.02
e
#4
Posted 17 February 2006 - 04:23 PM
Two comments.
It can be misleading since your sections are not even, thus the eye is drawn (at least my eye) is drawn towards the outer ring sections which are larger. But since your data in not ordinal this may not be appropriate. All days are even in this analysis right?
Intuitively I also want to see each day as a column...have you tried dividing your pie into seven sections with 24 divisions inside? (i.e. the reverse). probably not worth it...
It can be misleading since your sections are not even, thus the eye is drawn (at least my eye) is drawn towards the outer ring sections which are larger. But since your data in not ordinal this may not be appropriate. All days are even in this analysis right?
Intuitively I also want to see each day as a column...have you tried dividing your pie into seven sections with 24 divisions inside? (i.e. the reverse). probably not worth it...
#5
Posted 17 February 2006 - 08:05 PM
i'd second martin's comment about the different sizes for different days. plus i'd mention that i was not at all getting what the 0-23 where for until I saw Polaris' clock analogy comment. Perhaps a reworked title that also mentioned hour of day in addition to day of the week. I think if you made a bar chart with days on the Y, hour on the X, and quanity with a color variable you might just nail it.
good luck.
rob
good luck.
rob
#6
Posted 18 February 2006 - 02:17 AM
I think I would prefer a bar chart... One almost have to seem two alternatives next to each other to choose....
#7
Posted 19 February 2006 - 01:02 PM
Thanks for all of your comments. I must admit I had the same misgivings as Martin re the different size of the sections for the different days. I do like the clock analogy though, but one of our Crime Analysts thought that it was over-complicated and that our police officers wouldn’t understand it compared to the type of bar-charts they currently use!
Greg Driver
GIS Analyst
MapInfo User...!
#8
Posted 20 February 2006 - 07:43 AM
Very creative graphic, I really like the idea.
Make seven clocks and the officers will get it instantly how to read it.
My main point is, that you are violating Bertin´s rule on colour use.
According to Bertin, if you want to express real/interval scaled data (as the exact number of burglaries is) you are only allowed to use size as a graphic variable, not hue (colour) or brightness.
If you want to faithfully display classified/ordinal data, you may do so with brightness.
What you are doing is using brightness AND Colour. Colour change is suggesting a qualitative change. Like the change from arid to humid climate or freezing temparatures vs. warmer temperatures.
BUT: As my rant on another thread tried to point out, IF you actually do not want to portray the data faithfully, or you are told that your last and second last class have to stand out as of being of a different quality, than your decision was right.
Usually though I´m a big proponent of using only one colour in a graphic like this.
No need to use two graphic variables when the data changes only one dimension.
Make seven clocks and the officers will get it instantly how to read it.
My main point is, that you are violating Bertin´s rule on colour use.
According to Bertin, if you want to express real/interval scaled data (as the exact number of burglaries is) you are only allowed to use size as a graphic variable, not hue (colour) or brightness.
If you want to faithfully display classified/ordinal data, you may do so with brightness.
What you are doing is using brightness AND Colour. Colour change is suggesting a qualitative change. Like the change from arid to humid climate or freezing temparatures vs. warmer temperatures.
BUT: As my rant on another thread tried to point out, IF you actually do not want to portray the data faithfully, or you are told that your last and second last class have to stand out as of being of a different quality, than your decision was right.
Usually though I´m a big proponent of using only one colour in a graphic like this.
No need to use two graphic variables when the data changes only one dimension.
0 user(s) are reading this topic
0 members, 0 guests, 0 anonymous users


Sign In
Create Account

United Kingdom
Back to top
Netherlands
United States
Sweden
Germany








