Discussion


Note: Images may be inserted into your messages by uploading a file attachment (see "Manage Attachments"). Even though it doesn't appear when previewed, the image will appear at the end of your message once it is posted.
Register Latest Topics
 
 
 


Reply
  Author   Comment   Page 3 of 4      Prev   1   2   3   4   Next
cherdarchuk

Registered:
Posts: 29
Reply with quote  #31 
I actually find the sequence quite intuitive (the ring size does not change).  Here is your number suggestion with numbers greater than ten. When they overlap it is confusing, but it works quite well with no overlap.

Attached Images
Name: concenlegend.png, Views: 155, Size: 4.36 KB

Name: numbers2_o.png, Views: 152, Size: 30.76 KB

Name: numbers.png, Views: 155, Size: 59.50 KB


sfew

Moderator
Registered:
Posts: 823
Reply with quote  #32 
cherdarchuk,

Now that you've shown the full sequence of bubbles, I can distinguish them when they are arranged side by side, but they are difficult to discriminate and decode when displayed on a map. I get what you're trying to do, however. You're trying to make it easier to see small differences in bubble sizes by adding a dot in the center to make slight differences in bubble sizes for intervals 1 through 4 a little easier to see, and by adding rings to make slight differences in bubbles sizes of intervals 5 through 9 a little easier to see. This is enabled by the fact that the distances between the dot in the center and the perimeter of the bubble (i.e., the radius') differ by a greater percentage than differences in diameter or area, and by the fact that distances between the ring and the perimeter of the bubble also varies by a greater percentage than the difference in diameter or area. The problem is that by adding dots and rings, you have added visual complexity to the bubbles that creates visual clutter and slower processing as a trade off for improved precision. Because this greater precision cannot be perceived preattentively, but requires slower attentive decoding and comparison, I don't think enough benefit has been gained to make up for the extra complexity that has been added.

__________________
Stephen Few
cherdarchuk

Registered:
Posts: 29
Reply with quote  #33 
A quick correction, the image I included previously accidentally excluded the first bubble.  I think we'll just have to disagree on whether the benefit outweighs the extra complexity introduced by the white bands in the bubbles.

Attached Images
Name: concenlegend.png, Views: 150, Size: 4.51 KB


sfew

Moderator
Registered:
Posts: 823
Reply with quote  #34 
cherdarchuk,

It is certainly possible that your ringed bubbles would be more useful than I imagine. I'm only making a judgment based on my knowledge of visual perception and cognition, but it's too close to call for a definite verdict. In this case, neither your opinion or mine ultimately matters. Only empirical tests can determine the effectiveness of this design.

__________________
Stephen Few
cherdarchuk

Registered:
Posts: 29
Reply with quote  #35 
Agreed. Just a note that these are not my banded bubbles (the name I think I like best), credit goes to Francis Gagnon (who should also get naming rights).
sfew

Moderator
Registered:
Posts: 823
Reply with quote  #36 
cherdarchuk,

Does Gagnon's design match what you've been illustrating or is your design significantly different?

__________________
Stephen Few
cherdarchuk

Registered:
Posts: 29
Reply with quote  #37 
His latest version matches what I illustrated (minus the dot at the center which was an artifact of my implementation that I decided to leave in but is not really necessary unless precision in location is desired).
danz

Registered:
Posts: 190
Reply with quote  #38 
Cherdarchuk,

I expressed already in this topic my appreciation for the way this method handles the overlapping, but also my concern related to the size of the concentric circles. Too many concentric circles would lead to a a large size quite fast.
As also Stephen mentioned, I am tempted to "count" the circles, slowing down the comparison process.
Another downsides of any of the circle methods I saw till now is the difficulty of visually comparing of, for instance, 10 different sizes. 

However based on an interesting approach found on your blog, a simple idea came up in my mind which removes all the above mentioned issues.
1. Size
2. Counting temptation
3. Comparison difficulties for larger range of values.

Below a sequence of 10, which can be easily grown to 15 without any problem.

Dan



Attached Images
Name: Concentric_Bands.png, Views: 154, Size: 3.74 KB


cherdarchuk

Registered:
Posts: 29
Reply with quote  #39 
Dan,

Here is a rough go at what your method would look like.  It feels confusing to me, like we are plotting different categories not a continuum.  There is no sense that 5 and 6 are quite close in value.  




Attached Images
Name: danz.png, Views: 129, Size: 37.86 KB


danz

Registered:
Posts: 190
Reply with quote  #40 
Cherdarchuk,
 
Thank you for the effort of making my proposal available to this discussion. The effect I see on the map looks even better then I thought. You may also try 3 ranges of 3, 3x4 or 3x5 values. However, the values of radius for second range values can be just 1.5x instead of 2x of the radius of first range. 
 
The effect is exactly what I wanted to achieve. To use ranges of values rather then continuous values. 
Is what we usually do with numbers as well, no? First we compare the magnitude (amount of figures) and then we compare the values within the same order of magnitude. What is the most difficult visual comparison below?
 
  999,999
  999,998
 
  999,999
1,000,000
 
1,999,999
2,000,000
 
So indeed, the downside of my proposal is that we cannot say that the difference between 5 and 6 is the same as between 6 and 7. But is no confusion in comparing 5 with 6 or 6 with 7, right? The repetitve pattern of small amount of values 5,4,3 on a different scale improves a lot the comparison process. Another downside might be that the "area" of the shapes might be not in ratio with the values (still have to do some math here). But we perform quite bad in comparing areas, don't we?

The proposed method has the advantage of using just visual size comparisons, with no counting at all. With a small amount of values per range (3-5) I can easily identify the next shape in range, the first and the last shape in range even the middle one. With 2 or 3 ranges of 5 values the visual comparison effort is considerable lower then for one range of 10-15 values. Ranges of values might provide a good visual grouping. Eventually, the method can be easily used for arbitrary interval values, not only for linear or logarithmic scale. (Is it an idea to call it ranking encoding rather then quantitative?)

Dan
Tim2

Registered:
Posts: 57
Reply with quote  #41 
danz, I don't get this at all. You are using two attributes (bubble size and degree of bubble fill) to visually encode a single variable. I can't, personally, see this being helpful.
jlbriggs

Registered:
Posts: 194
Reply with quote  #42 
Danz - I also have to say, I get nothing but confusion from looking at that on the map, and would much prefer to see standard bubbles to encode the data.

I would say that it decidedly adds confusion instead of removing it.
danz

Registered:
Posts: 190
Reply with quote  #43 
It is obvious that there are more negative votes then positive. I absolutely agree that this counts the most.

As I said before is just a variation of an existing method (this reference and this reference), was nice to see it on a map, so me and others could express an opinion about it. 

@jlbriggs: in terms of visual cumulation, simple bubbles perform the best, even if they overlapp. Visual comparison stays difficult for 10 and up different sizes of bubbles.
 
@tim2: you pointed right that they are 2 attributes. the external diameter encodes the range and the inner one the position within the range. This actually helps to do the comparison faster then in any of the other cases.

This method fails to be a solution for a map mainly because on a map we focus on cumulative effect, precise comparison based on an uniform incremental scale rather then just simple ranking. However, for ranking and grouping purposes, this encoding method looks superior to the other methods which involves counting or small variations of one dimension.
jlbriggs

Registered:
Posts: 194
Reply with quote  #44 
"However, for ranking and grouping purposes, this encoding method looks superior to the other methods which involves counting or small variations of one dimension"

I have to say I still just disagree across the board.

Seeing the rings laid out in order is still confusing, with the added danger that when you judge the size of a circle wrong, you are now getting it wrong by a whole order of magnitude.

And you've spent more cognitive energy to get it wrong.

I think it was a great idea to test out.  I just don't see it working, on or off a map.
danz

Registered:
Posts: 190
Reply with quote  #45 
When I saw this method first  (this reference and this reference), I definitely could see the comparison between donuts better then between bubbles. 

For me the sequence is not confusing. I don't have issues in visually comparing, grouping or sorting. The outer diameters of the donuts are more distinct then the diameters of the bubbles. I can compare better two donuts with similar sizes either being close or far from each other within a map. But, overall, I don't like the way the donuts perform on a map.

They are indeed many issues with this method to be ever considered: the broken sequence, two dimensions, no scale comparison, missing cumulative effect etc. 

Below is a more precise representation of a scale from 1 to 10 for donuts and bubbles. When I say precise, it means the area of the shapes is in ratio with numbers from 1 to 10 for both representations.

The main subject of this thread was Stephen Few article about bricks. It changed quickly into a discussion about encoding the maps in a more efficient way. It seems that "the bubbles" are still the best answer to this matter.

Attached Images
Name: Bubbles_vs_Donuts.png, Views: 107, Size: 3.56 KB


Previous Topic | Next Topic
Print
Reply

Quick Navigation:

Easily create a Forum Website with Website Toolbox.