Multiple regression: one of my significant factors has outliers, what should I do?

#1
So I ran a multiple regression and found two factors that explained one of my dependent variables quite well. The problem is that one of the two factors has what I guess you could call outliers:

http://i.imgur.com/7h7NcjD.jpg

Note that three of my points are on 0 on the x-axis. Is this a problem? Should I drop them from my data set? I'm wary since I already have such few data points. What do you guys think?
 

kiton

New Member
#2
Re: Multiple regression: one of my significant factors has outliers, what should I do

Hello!

Six data points won't do much for you at all. Yet, for practice you can calculate DFBETA and/or Cook's distance to examine how influential outliers are.
 
#3
Re: Multiple regression: one of my significant factors has outliers, what should I do

I agree with the above. With only six observations your regression model is considerably compromised.

However, if you nonetheless would like an answer, I would say no, you should not remove them. At least not if there is no reason to believe that the values are wrong. If there on the other hand upon closer inspection is reason to believe that measurement error has occured, then removing them would be reasonable.