You still need to do the calculation in the correct way. Using the average here might make more sense intuitively for this outlier but if you get into that habit you’ll wind up making a lot more mistakes on those that aren’t outliers.

What you can do is calculate both and when they differ so greatly you know there’s some volatile data that might bear more investigation and explanation.

]]>My problem is calculate CTR (Click-through rate : # clicks / # views) for a period

Let’s say we have this situation:

Clicks : 1 1 1 1 1 1 1 1 1 1

Views : 1 1 1 1 1 1 1 1 1 100

Ideally, without the the last element(it’s probabily a outlier),we will say that the CTR is 1 right? But

Mean(of CTR by day) : 0.90100

Sum(clicks) / Sum(views) : 0.091743

In this situation the mean of means best fits our intution, because it’s limits the effect of outlier.

What do you think?

PS : Sorry for my bad english.

]]>My particular instance: I am trying to report on the average search impression share lost to budget. A simple average doesn’t take into account the weight of each number, but what, most appropriately, does?

I would guess I’ll have already made a choice and moved forward by the time someone has an opportunity to respond, but I’ll try to remember to check back in to read your thoughts.

Take care.

]]>Sum this column and divide by the sum or all impressions. This number is your true Average Position.

]]>