| Date: | Tue, 16 Jan 2001 11:26:49 -0500 |
|---|
| Reply-To: | Arthur J Kendall <KendallA@GAO.GOV> |
|---|
| Sender: | "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU> |
|---|
| From: | Arthur J Kendall <KendallA@GAO.GOV> |
|---|
| Subject: | Re: When comparing two group means.. |
|---|
|
| Content-Type: | multipart/mixed; |
|---|
It all depends. Some of the considerations are the answers to these questions?
Why are these cases outliers on these variables? Are the values obviously nonsense? ( I.e., are they outside the theoretical range of the measurement)? Very commonly "outliers" are data entry errors. Have you key verified or proofread your data?
Can you go back to the data gathering phase and get these measures again?
Do your conclusions change when you include vs exclude these cases?
How far out are these cases in Z=scores? Are they in both directions from the means?
What is the nature of the variables and why are the particular values extreme?
Are these cases univariately, bivariately, or multivariately unusual?
>>> "Luo, Peter" <pluo@DRAFTNET.COM> 01/16/01 09:58AM >>>
Both groups have a small number of outliers which affects the group mean
dramatically (sample size is large). Should I remove them before I conduct
the comparison in order to get a more accurate reading?
****** Message from InterScan E-Mail VirusWall NT ******
** No virus found in attached file noname.htm
* All attachmes were scanned by Interscan 3.5, and no virusus were found.
***************** End of message ***************
| 
