David Heiser wrote: > > -----Original Message----- > From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]]On > Behalf Of Glen Barnett > Sent: Wednesday, March 20, 2002 4:26 PM > To: [EMAIL PROTECTED] > Subject: Re: Coef. of skewness - Is my prof. crazy? > > David Heiser wrote: > > Fisher > > changed things, and now we measure skewness using Fisher's method. The > Excel > > method is the correct Univariate way. > > In what sense is the measure you refer to more "correct" than another? > > Glen > ---------------------------------------------------------------------------- > --- > Let me ask you a related question: > Which is the correct method to estimate the variance (i.e. sigma squared) of > a normal population from a sample of N measurements (values) from this > population? > a. The maximum likelihood method (i.e. the (sum(xi-xbar))/N > b. The unbiased estimate (i.e. the (sum(xi-xbar))/(N-1) > And you are assessing a software package for correctness in reporting a > variance result?
To say one of these is "correct" and the other is not is nonsense. Both are sensible estimators. They have slightly different properties, and you choose one based on what you need your variance estimate to do. If the software package does not *say* which it is using (very naughty), it is easy enough to figure out by giving it the sample (0,1). You can then decide if it is suited to your purpose, and if not, adjust its output.* As long as it calculates one of the two correctly, it is correctly reporting a variance. *The conditions under which /n or /(n-1) makes any real difference are rare at best. Glen . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
