Click here to see the SAS code.
Click here to see the example.

Click here to see the contest scenario description.

---------------------------------------------------------------


DM Review Magazine Contest - Scenario 2:

Approach:

For this scenario, I chose to use a custom "scatter plot".

Markers:

For each employee, I place a marker at the corresponding salary*salary_grade,
but instead of placing it _exactly_ at the salary grade I offset the males
slightly to the left, and the females slightly to the right, so that they
can be compares & contrasted more easily.  To reinforce the distinction 
between male & female markers, I also color the markers blue & pink 
(blue=male, pink=female).  

I draw a subtle/gray box at each pay grade, showing the min/max of the
target salary range for that pay grade.  You can easily see whether
the markers fall within the box or not.  

Also for each pay grade, I place special pointer/markers, denoting 
the average male and female salaries for that grade.  This gives an
easy way to compare the male and female salaries (just looking at the
markers alone could be misleading, because there would be no way to 
detect multiple markers that exactly overlap).

Legend:

The legend for the individual markers, the average pointer markers,
and the target salary grade range was placed inside the plot, because
there was plenty of room (there should always be room at the top/left
of this chart, since the salaries for the low grade range will 
always be lower).

Axes:

Maybe the most important axis feature is the calculated %-inequity,
which is printed for each salary grade, along the x-axis.  This is very
important - without this %-inequity, people might be tempted to just 
look at the salary difference between the male & female average, 
but by looking at the %-difference you get a better and more true
comparison.  I made these %-inequity numbers larger/bolder than the 
salary average labels, to show that they are more important.

I angled the y-axis label, because it seemed to look better (less
cluttering) there than at the top of the axis.  I also placed gray
dashed-line reflines at the y-axis tickmarks, to you can easily tell
which markers are above/below these $20k salary milestones.

Observations/Results:

Although the salaries of most of the employees (both male and female) 
in each salary range are within the 'target salary' range (boxed-in area), 
it does appear that there is a discrepancy between male and female 
salaries in this company.  This salary discrepancy is most pronounced
in the lowest (1) and highest (5) salary grades.  This should warrant
a more detailed analysis, including other factors such as years-of-service, 
the specific job being done, etc.

Worth mentioning:

Although not helpful in a printed copy or static gif image...
the html-web version of my output allows you to hover your
mouse over each individual salary marker and see the employee number 
and value in the charttip (aka mouse-over text, rollover text, alt text).
I can send you the html overlay upon request.


Here is a link to the article describing the winning entry.
They had a very similar layout as mine, but they used a 
box-and-whisker plot (imho, I'm not sure that 5 data points
per group is enough to do a valid box plot.) ...

Winning Chart
  

Back to Samples Index