0

I have 12 instances for the customer usage i.e 1200,1220,1230,1190,1250,1180,1270,1210,1240,1300,1290,1280.

The simplest way to calculate average is No.of Instances/12. But if sometimes one of the instance has abnormaly high value e.g. 6000. It causes the overall average to deviate greatly.

What is the best way to cater this kind of situation without excluding the high number?

I know that one solution could be to define a certain threshold and to compare the value of each instance before the average calculation and the other solution could be to take into account the standard deviation. But both of these solutions somehow are not feasible for my data.

If anyone has a better solution then please share. Thanks

user3734553
  • 3
  • 1

1 Answers1

1

The classical solution to this kind of problem is using the median instead of the average. You can also combine these two ideas: discard a certain amount of values (e.g. the upper and the lower 25%) and compute the mean of the remainder. This is called a truncated mean.

MvG
  • 57,380
  • 22
  • 148
  • 276