home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!sun-barr!ames!agate!violet.berkeley.edu!lele
- From: lele@violet.berkeley.edu (Sharachchandra Lele;100 BldgT-4;26886;5106440692;NT23)
- Newsgroups: sci.math.stat
- Subject: Bug in CSS: Statistica program?
- Date: 30 Jul 1992 05:40:48 GMT
- Organization: University of California, Berkeley
- Lines: 20
- Sender: Sharad Lele
- Message-ID: <157vd0INN374@agate.berkeley.edu>
- References: <1992Jul3.163117.9940@jupiter.sun.csd.unb.ca>
- NNTP-Posting-Host: violet.berkeley.edu
- Summary: Standard error calculation is "wrong"
-
- While reviewing CSS:Statistica (for DOS) (a summary of which I posted
- recently) I found the following "Bug" that I would like to crosscheck
- with others who have used this program:
-
- CSS uses the "sample std. dev." formula for calculating standard
- deviation. However, when it calculates standard errors (either in the
- ANOVA-descriptive stats command, or in GRAPHICS:box/whisker plot), it
- appears to use the "population s.d." value! I noticed this when I was
- cross-checking the results using SYSTAT.
-
- As far as I know, the convention is to use sample s.d. always.
- Have others noticed this bug? Or am I making a mistake? or is there
- nothing wrong with this method?
-
- --Sharad Lele
- Energy & Resource Group
- Univerity of California, Berkeley
-
- Electronic-mail: lele@violet.berkeley.edu (Internet)
-
-