From: Brian Hawkins <hawkins@goasa.org>
To: Teacher2Teacher Service
Date: Dec 15, 2007 at 22:38:21
Subject: Outlier detection

I was looking for a simple way to detect outliers in bivariate data and I 
realize outliers can come in many forms when you look at a scatter plot.  
I was thinking that I could have students create a residual plot using 
their line of best fit and then use the residuals to calculate outliers 
based on 1.5 times the interquartile range. It seems like this could 
work although I came upon a source that said it would be better to square 
the residuals. It also seems like this possibly could detect outliers in 
the y values. I only thought of this because it could incorporate 
multiple skills that my students should already have. Any ideas if this 
is feasible or if there are other simple ways that don't involve college 
level math or AP statistics level math?

Post a public discussion message
Ask Teacher2Teacher a new question