home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!stanford.edu!bu.edu!olivea!spool.mu.edu!howland.reston.ans.net!paladin.american.edu!auvm!UICC.COM!SDH
- From: sdh@UICC.COM (sheldon haynie)
- Newsgroups: bit.listserv.stat-l
- Subject: (no subject given)
- Message-ID: <9301261700.AA02710@procad.noname>
- Date: 26 Jan 93 17:00:42 GMT
- Sender: STATISTICAL CONSULTING <STAT-L@MCGILL1.BITNET>
- Lines: 39
- Comments: Gated by NETNEWS@AUVM.AMERICAN.EDU
-
- >>Hi, STAT-Lers!!!
- >>
- >>Can anyone give me a reference on this outlier test?
- >>
- >>Q = ABS(outlier - mean) / ABS(max - min)
- >>
- >>I haven't been able to find this in any ASTM procedure, QC book, or stat
- >>book. I suspect that it may be an "engineer's" formula but I thought I'd
- >>check with the experts first. TIA.
- >>
- >>Janet P. Buckingham
- >>Southwest Research Institute
- >>6220 Culebra
- >>San Antonio, Texas 78228
- >>210-522-2407
- >>
-
-
- I can't give a formal answer regarding your *Q* test...
- but as an "engineer," who often deals with noisy data if would seem
- like a typical statistical *abuse* by quantifying the spread of data
- with outlier's vs without... Ideally you would want a low value of
- Q (.5 in limit ?)
-
- I have often pre-conditioned my data this way to avoid misinterpretation
- due to instrumentation errors...
- (signal to noise and all that.) Of course you need to not filter out
- the signal when identifying "noise"
-
-
-
-
-
- Sheldon Haynie |
- Unitrode IC Corp |
- 7 Continental Blvd. |
- Merrimack, NH 03054 |
- FAX:603-424-3460 |voice:603-424-2410 (EST)| sdh@.uicc.com
- *********************************************************************
-