Search All of the Math Forum:
Views expressed in these public forums are not endorsed by
NCTM or The Math Forum.



Re: Ftest for comparing variances?
Posted:
Nov 9, 2006 5:20 PM


Richard Ulrich wrote: > > There is an interesting little point about Fdistributions at the > end... > > On 6 Nov 2006 19:42:57 0800, "Reef Fish" > <large_nassua_grouper@yahoo.com> wrote:
> > > I'm pretty sure that the "larger over the smaller" was the old > > > convention. I stated that, and how to use it; and Art restated it. > > > > There is NO such convention. You should have read what I posted. > > Bob also writes his posts without reading the next sentence.
< numerous gratuitous and irrelevant comments by Richard Ulrich snipped> > > [snip] > [snip] > > > > > RU> For some extreme cases, that formula gives a value greater than > > > RU> 1.0. To avoid that, or to use a computer program for it, you can > > > RU> get the exact pvalue and if it is greater than 0.50, subtract from > > > RU> 1.0; then double it. > > > > And I asked the question: Probability greater than 1? > RU> SPSS used to provide the "Fratio test for variances" > RU>  I was amazed when I first saw pvalues reported as greater RU> than 1.0, like "1.07". > Richard was now trying to blame SPSS for his OWN errors.
> > From Google, < define:gratuitous > > > Definitions of gratuitous on the Web: >  without cause; "a gratuitous insult" >  complimentary: costing nothing; "complimentary tickets"; "free > admission"
As I had said to Richard, any insult on HIS error of getting a probability of greater than 1 is hardly an INSULT, let alone gratuitous.
 Reef Fish Bob.



