Best way to answer average question with large range of data I have expense data for 30+ departments. I want to figure the best way to answer the question 'what is the average expense?' The problem is that each department has a different range of data and size. If I take the average of everything the answer will be skewed. 
Sample data:
dept A frequency: 8
dept B frequency: 19
dept C frequency: 28

dept A: 1, 2, 3, 4, 1, 2, 4, 5
dept B: 30, 40, 20, 25, 30, 40, 40, 40, 50, 30, 35, 60, 30, 50, 55, 45, 43, 33, 32
dept C: 1000, 2000, 1500, 3000, 3240, 4000, 2300, 1000, 2000, 1500, 3000, 3240, 4000, 2300, 1000, 2000, 1500, 3000, 3240, 4000, 2300, 1000, 2000, 1500, 3000, 3240, 4000, 2300

dept A average: 2.75
dept B average: 38.32
dept C average: 2434.29

Some ideas I have are to take the average of averages, give an average for different bins of expenses (this will remove the department level), normalize the expenses, or standardize the expenses.
I am open to any suggestions on how to communicate a good answer to the question.
Thanks!
 A: Given the data in your example, it makes sense to report department averages as well as the overall average (as if the data were all coming from one department) and then provide commentary around why the overall taken by itself hides important information.  Showing the department averages puts the spotlight on Dept. C!  This will answer the question but also protect against misinterpretation.  
Note:  If you report the average of averages, that result will give each department equal weight in the calculation and this may be undesirable.  In particular, going that route masks the fact that Dept. C has more than 3 times the amount of data relative to Dept. A (in addition to bigger numbers).  If each department was reporting the same number of data points, the overall average would equal the average of the department averages - which isn't the case here.
If you also have bins of expense, you could generate a table, such as the one below, for each bin.  But then there are lots of other ways to cut the data at that level.  

