Regression: Module #3–The impact of an outlier

Module #3, The impact of an outlier

Download PDF Copy Go to Regression Applet

This problem is a little different than the others. Again, if you don’t have a copy of this page you should print it. If you have a copy of the page, go directly to the applet.

For this problem, first distribute the data points so that you have a correlation of about +.90.

A.  Imagine that you added an extreme point to the upper left-hand corner (don’t actually add the point yet). How do you think this would change your correlation?

B.  Would you expect the correlation to become larger or smaller? ______________

C.  If you had to guess, what do you think your correlation value would be after the addition of this single point? ___________

D.  Now add a point to the upper left-hand corner (to add a point, double click the spot where you want to add the point). This new point will represent a single point that deviates from your strong pattern of correlations.

E.  What is your new correlation? ___________________

F.  Does this differ from your expectation? How? Comment on the impact of the addition of an extreme score. (It may be interesting to look at the squares for errors – Check “Show SS error” and “Show Error as squares.”)

G.  Imagine that you had a set of scores where one value was incorrectly entered (e.g., X = 100 entered when x was actually 10). How might this type of error impact your results?

Begin Module 4

667 total views, 1 views today