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  
BillMedrano

Registered:
Posts: 2
Reply with quote  #1 
All:  Bill Medrano here, a big fan of this work and quite grateful for the insights, coaching, and direction it's given my peers and myself.
 
Using a dual axis chart, I plotted out fuel prices and vehicle sales over time, distinguishing Domestic and Foreign vehicle origins.   
 
My question is that since we are seeing the effect of Fuel Price on sales, we aren't comparing magnitude.  Consequently, I used a black bar mark type for fuel price to suggest a background to show vehicle sales against.
 
Thoughts? 
 
Thanks in advance!
 
Bill Medrano
 

Attached Images
Name: Fuel,Sales_over_time.jpg, Views: 88, Size: 66.06 KB



__________________
Bill Medrano

jlbriggs

Registered:
Posts: 190
Reply with quote  #2 
My thoughts on the dual axis are pretty straightforward:  if the two axes are measuring two different incomparable units, the different series don't belong on the same chart.

In my opinion, there is no positive result of having the fuel price series plotted on the same chart as the car sales series that can't be provided by added a second chart below the car sales chart that plots the fuel prices.

On the other hand, there are plenty of potential negatives, primarily the visual clutter that it adds, and the potential for misleading comparisons caused by the scaling of the axes and the interaction between the fuel price plotting and the car sales plotting.

Often these comparisons are automatic - you look at the point where the orange line dips into the black bars as if it means something, even if logically you know it does not.  And plenty of users will *NOT understand that it does not have meaning, to boot.


That aside, I also see your title as misleading.

"Effect of" anything on anything else is not demonstrated by this chart.

Simply calling it "Fuel Prices and Vehicle Sales" would be more appropriate.
danz

Registered:
Posts: 181
Reply with quote  #3 
Bill,

Correlation does not necessarily imply dependency. Effect of fuel price on sales volume is an assumption based on common sense, not on mathematical dependency. 

I used in past myself dual axis representations similar with yours, but not anymore. It was always an issue in establishing the independent scales per graphs and then to "correctly" overlap them. Nothing stops you to use for fuel costs axis limits 0-5.0, so graphic will be as "correct" as before even if it will look different. 

To have a consistent correlation graph over time for measurements of different unit of measures, define a uniform unit of measure which might be % of maximum value. Then use only line graphs for all 3 measurements and you have the right representation. You can use for Domestic and Foreign a common maxim (% of max(Domestic, Foreign)) or you can use % of maximum value for each measure independent (% Max(Domestic), % Max(Foreign), % Max(Domestic+Foreign), % of Max(Fuel Price)).
BillMedrano

Registered:
Posts: 2
Reply with quote  #4 
Thank you Danz - that makes a lot of sense!

Much appreciated!

Bill

__________________
Bill Medrano
wd

Registered:
Posts: 167
Reply with quote  #5 

The only correct display for correlation is an xy scatter plot...


__________________
Bill Droogendyk
danz

Registered:
Posts: 181
Reply with quote  #6 
A scatter plot makes sense for bivariate analysis, indeed. Different type of relationships can be identified visually from such of graph, not only linear correlation. However in above example time was an important aspect and a scatter plot is not able to easily display that (connected points in the order of events or increased in time color saturation might help a bit, but not much).

If we study the variation in time of several measures within their own range or against their own targets - calculated in percentages, we are able to have a consistent display in a common chart. 
Previous Topic | Next Topic
Print
Reply

Quick Navigation: