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
sfew

Moderator
Registered:
Posts: 803
Reply with quote  #31 
Derek_C,

Your inversion of the "on time" values to "not on time" works nicely. The one thing that ought to be changed, however, is that the red line which encodes the "not on time" values should be displayed as zero on the graph when the value is zero, rather than not appearing.

__________________
Stephen Few
Derek_C

Registered:
Posts: 69
Reply with quote  #32 
Yes, I don't know what was happening there. I suppose it's to do with Excel's lines appearing at the "bottom" of the stack of chart objects, whiile its symbols appear at the "top". Since I had no symbols, the lines disappeared. If I'd been prepared to spend more time I think I'd have jittered them very slightly to see if they would appear above the scale line.
JonPeltier

Registered:
Posts: 6
Reply with quote  #33 
Shortly after Stephen's blog entry about dual axis charts, I replied with general agreement that secondary axes are generally detrimental to the clarity of a chart:

http://peltiertech.com/WordPress/2008/03/25/secondary-axes-in-charts-2/

I noted that the secondary axes in Excel can be helpful when constructing effects that otherwise seem impossible. I also presented a panel chart, which is a way to combine two or more related charts into separate "panels" within a single Excel chart.

Soon after posting this, I remembered that a secondary scale for a single data series which may have multiple measurement units (Fahrenheit vs. Celsius, miles vs. kilometers, etc.) is actually a valid use for a secondary axis which adds value to the chart:

http://peltiertech.com/WordPress/2008/03/26/secondary-axes-that-work-proportional-scales/

I regret missing this forum thread until now, but I am pleased to see that Naomi and Stephen have both validated this use for a secondary axis.

__________________

- Jon
-------
Jon Peltier, Microsoft Excel MVP
Tutorials and Custom Solutions
Peltier Technical Services, Inc. - http://PeltierTech.com
_______
khailey

Registered:
Posts: 1
Reply with quote  #34 
grasshopper: Liked your Fed Ex graphs - worked well for me.

One thing no one seems to be discussing is compacting data, ie reducing the real estate it takes to convey the data. Compacting data, in an understandable form, is one of the primary reason to use graphics. Graphics can take pages and pages of text and convey the information in a condensed form that is consumable with greater speed and less effort.

The main reason I would want to use dual axis graphs is to compact data and save real estate and flipping my visual area back and forth. If my visual area is compacted by collocating the data and the collocating doesn't add much confusion then it's seems like a good idea. The basic trade off comes from the gains from data compaction vs increased confusion between unwarranted correlations. The case for unwarranted correlations seems over blown in the discussion. I do agree, though, that dual axis graphs are to be used with discretion if at all.

grasshopper's second Fed Ex graph is a nice example of clarity and compaction with a dual axis graph

Just given the amount of discussion here, it's clear that it's not obvious whether dual axis graphs good or bad.

Kyle Hailey
http://sites.google.com/site/youvisualize/
jamoroch

Registered:
Posts: 1
Reply with quote  #35 
Though the thread is old, I would like to add this. I think there is a situation where these plots are useful. I work on numerical analysis where you are constantly measure the error between the "real" value and the simulated value. I have found useful to plot the real and simulated values on one axis and the error on the other. This leverages the information of the error. usually, the error is measured in a logarithm scale or as percentage.   

I would love to know  your thoughts on this matter.

sfew

Moderator
Registered:
Posts: 803
Reply with quote  #36 
Jamoroch,

If you provide an example, I and perhaps others as well will let you know what we think.

__________________
Stephen Few
mjoyce

Registered:
Posts: 11
Reply with quote  #37 

Hi,

I couldn't help but respond to this thread.

What of Pareto charts?

They use a second Axis.
I'm convinced Paretos have value, I use them regularly to understand what proportion of one is used in comparison to the top n other things.

For instance what proportion of effort is required to support the top 5 IT services, and what are the next services.

Regards
sfew

Moderator
Registered:
Posts: 803
Reply with quote  #38 
Hello mjoyce,

Generally, Pareto charts only include a single percentage scale, which applies both to the bars and the line. On those occasions when a Pareto chart displays two scales--one of the left (for example, dollars or counts) and one on the right (percentages)--they are actually two expressions of the same scale, not two distinct scales. The two expressions of the same scale apply to both the bars and the line. The dual scales that I warn against when using graphs to communicate information are those that are completely independent from one another, such as revenue in dollars on the left and a count of orders on the right, where some values in the graph are associated with one scale and other values are associated with the other.

__________________
Stephen Few
DataInAction

Registered:
Posts: 12
Reply with quote  #39 
I know this is an old thread but I was working on a project lately where dual-axis became a hot topic (again) so I thought I would share my experience.  This example is from a dashboard primarily used for monitoring.  Without going into too much detail the engineers who track this data are looking for occasions where the three values depicted on the chart are out of sync.  The absolute value of the data is also important, but not as important as the relationship between the lines.  The data is in the same units, but the values vary widely.  The first chart below shows the values as they were originally plotted on a single axis.


Note how the chart shows the three values, all with the same units over time (x-axis).  The values are the daily values plotted over approximately 90 days.  

In this case, there is not much of interest.  The blue line is difficult to compare to the red and yellow lines due to the scaling issue.  The red and yellow values are clearly rising, pretty much in sync, and it appears the values of the blue line is rising but this is harder to discern.  

The key point is that no issues jump out at me based on this display.  Below is the same data plotted on a dual axis:


From this graph the issue is clear.  For this one day near the center of the date range clearly something was wrong as the blue value was out of sync with the others.  While they appear perfectly in sync most of the time the exception now jumps out at me when I can see the data plotted together on different scales.  We ended up using dual-axis for all of these types of charts as the engineers felt they were the best way to visually locate anomalies.

As the original article discusses, there are many of examples where the use of dual-axis charts (especially bar charts) leads to displays and dashboards that are confusing at best, and deceiving at worst.  But sometimes the value of having all the data plotted together appear to make dual-axis charts the right choice.  I'm curious if anyone has ideas/concerns with using dual-axis in this way and/or any suggestions for improvements.  Thanks.




sfew

Moderator
Registered:
Posts: 803
Reply with quote  #40 
DataInAction,

Based on your description of its use and audience, this seems like a good example of a dual-scaled graph that works. It would work even better, perhaps, if you changed the background color to white and removed the grid lines, which would make it easier to focus on the data and spot occasions when the lines are out of sync.


__________________
Stephen Few
DataInAction

Registered:
Posts: 12
Reply with quote  #41 
Ah, yes, good point.  This is my quick Excel rebuild of the original chart but I looked at the original and it has the same problem (grid lines and background color).  Thanks.
wd

Registered:
Posts: 167
Reply with quote  #42 

Or, if the data really is this simple, multiply blue by 10 and indicate that fact with a label ...


__________________
Bill Droogendyk
PeterRobinson

Registered:
Posts: 33
Reply with quote  #43 
Perhaps show differences, ratios or a form of correlation since it seems the actual values might matter less than their similarity.

__________________
Peter Robinson
in Brisbane, Australia
Chiayulin

Registered:
Posts: 2
Reply with quote  #44 
I am leading a team whose main task is to create tons of dual axes charts to help the management or clients to understand the “interaction” or “relation” between two datasets or two factors which are often in different units.

However, because Excel’s current algorithm for scale setting only considers single vertical axis instead of both vertical axes, so in dual axes chart, the scales are not set in the same elongation, resulting that if using Excel dual vertical axes chart it would assume that there was no much relation between the 2 factors, and misleads to a wrong conclusion.

As I posted my questions on Excel forums, most suggestions I’ve got is that I need to manually adjust the scales of the axes every time to “create a chart that I think it makes sense”. However, manually adjust every charts is neither practical nor efficient for our work since we have a lot of charts and several team members while each team member should adopt the same scale setting method to create fair comparison.

Therefore, I've searched online to see if there is any solution could fix the problem we faced, and then I found a website with a patented algorithm which claims could solve the aforementioned problem, here I would like to share with you: 
http://www.graphician.com/patent-01.html

I've downloaded the add-in from the website and recreate the dual axes charts. To me, it seems that the add-in can do the correct calculation, and make the chart that I think it make sense. I would love to know how you think of this add-in, make sense to you? Thanks!
danz

Registered:
Posts: 182
Reply with quote  #45 
Chiayulin,

Are the dual axes charts referencing independent variables, each with its own scale? It looks like considering your remark: "“interaction” or “relation” between two datasets or two factors which are often in different units".

If this is the case, is no reason to use dual axis. If you need to display the variation in time of two or more variables you can use a common variation scale for the them. A recent post in this forum contains a debate regarding this matter.

PS.
Scaling algorithm patented? Seriously? I may not be aware with the meaning of a patent, but this is quite hilarious. On second thought, is not. Is sad.
Previous Topic | Next Topic
Print
Reply

Quick Navigation: