Detecting outliers non parametric data

#1
Hi,

what is the best (and easy) way to detect outliers in spss for skewed, not exactly normally distributed data?

I mainly need it for logistic regression an I assume that even tho logistic r. does not assume normality I still need to make sure that there are no extreme outliers

Thanks a lot
 
Last edited:
#2
Oh no, it is the same poster that made a double post on the other site.

(And I also made the reflection about "non parametric data". Where does these prejudices comes from?)
 

noetsi

Fortran must die
#3
Cooks distance applies to logistic regression as to regular regression. A value greater than one is sometimes seen as indication of a serious outliers. I do not know how SPSS, which I rarely use, deals with Cooks distance, but you might look that up.
 

noetsi

Fortran must die
#5
Be forwarned from my own experience that there are disagreements on what is the dividing line between an outlier and not in cooks's distance (technically this is measuring influence not just outliers I believe). One is just one of the values used and which value you do use can make a huge difference in what is labeled an outlier .