John Cook as made some excellent blog posts regarding outliers.  The point generally has been that we tend to underestimate improbable events when distributions have long tails.  (At least I think that might have been the point.)

I had a personal epiphany tonight with an outlier observation while I was preparing dinner.  I discovered couscous a few years ago.  I like to cook, and it has become one of my favorite side dishes.  My Mom and family never ate middle eastern food, and I do not recall ever being exposed to couscous until I was an adult that was able to purchase my own groceries.  The observation is that of all my genetic cousins, none of them like couscous but me.  I know there are a million very good reasons why I might like couscous, but looking at my family, I am an outlier when it comes to couscous.

My taste for couscous makes me a genuine outlier, maybe, probably?

Which leads me to my real point.  I am required to analyze reams and reams of data in my job as an analyst.  Unlike many industries and endeavors, I am covered up with data that would be the envy of many statisticians.  What I have learned from so much data, is that, it is what it is.  There are so many data outliers that it does not make sense to discard them.  Outliers occur often enough that they are normal.

It is important to learn to accept outliers in an analysis, rather than trying to exclude them from consideration.  Perhaps we should reflect inwardly on the state of our models.  Maybe they are not correct?