Skip to main content
Cross Validated

You are not logged in. Your edit will be placed in a queue until it is peer reviewed.

We welcome edits that make the post easier to understand and more valuable for readers. Because community members review edits, please try to make the post substantially better than how you found it, for example, by fixing grammar or adding additional resources and hyperlinks.

4
  • 11
    $\begingroup$ In Box, Hunter & Hunter: "Statistics for Experimenters" they tell that, in the chemical industry, outliers often have led to new patents. Do you want to throw out your new patent? $\endgroup$ Commented Mar 8, 2016 at 20:40
  • 3
    $\begingroup$ Nope, I don't want to miss out on any patents. But I also don't want to spin twelve cycles trying to get my model to accommodate "somebody pulling on the wires." That's almost certainly not the phenomenon under study. I do like the idea of outliers as opportunities, and one thing to be said for straightforward deletion is that at least the code will provide documentation of those deletions, whereas in robust methods the outliers just kind of coexist with the other points. $\endgroup$ Commented Mar 9, 2016 at 14:26
  • 3
    $\begingroup$ You are right that the specific circumstances must be taken into consideration. What should not be done is apply some context-free "rules" for outlier rejection. There do not exist any such good rules. $\endgroup$ Commented Mar 9, 2016 at 14:36
  • 1
    $\begingroup$ My favorite point about the power of context is illustrated by the question, "Are Snickers bars healthy?" Well, if you've been lost in the woods for three days and you've just found a few on the ground, it turns out they're pretty healthy after all. I feel like the popular answers here are telling us, "Never eat a Snickers bar, unless you're absolutely sure you'll die if you don't." $\endgroup$ Commented Mar 13, 2016 at 21:07