Skip to main content

Hi,

I have a scenario like this -

 

  1. 100 people with say an average score of 53.

 

  1. Let’s say I slice this 100 people to find out the number of managers and their score, I get 10 managers with an average score of 58 and the non managers are 90 with an average score of 53.

The averages of 53 and 58 is 55.5 ( rounded off to 56 )

 

Now the difference is a delta of 3 between the full population & the average of the subpopulations ( 53 vs 56 ).

 

The above scenario is something I see when I analyze the recent engagement survey. Could some one explain why the average scores of the sub populations do not match the average score of the full population ?

 

@vipin.shyam this is a questions that support@glintinc.com can help you with!

Best,

Judy


For the benefit of people in the community, let me attempt answering my own question. 

The average of averages is not a recommended check. Why ? Below is why :

 

In the below instance the combined average is 4.375. If I were to take an average of only employees it would be 1 & if I were to take average of managers separately it would be 10. The average of averages in that case would be ( 1 + 10 ) /2  = 5.5, which is different from the overall average of 4.375.

 

Category Score
Employee 1
Employee 1
Employee 1
Employee 1
Employee 1
Manager 10
Manager 10
Manager 10
Average 4.375

 

 

The only time average of averages work is when each of the subpopulation has exactly the same number of people. For example, I increase the number of managers to match the number of employees as below :

 

Category Score
Employee 1
Employee 1
Employee 1
Employee 1
Employee 1
Manager 10
Manager 10
Manager 10
Manager 10
Manager 10
Average 5.5

 

Here the combined average is 5.5. The average of employees is still 1, the average for managers is still 10, however, the average of averages ( 1 + 10 ) / 2 = 5.5  which is now equal to the combined average of 5.5.

 

Hope this clarifies.

 


😣

But thank you!


Reply