Date: Mon, 4 Jun 2001 10:33:46 +0100
Reply-To: "R. Allan Reese" <R.A.Reese@GRI.HULL.AC.UK>
Sender: "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From: "R. Allan Reese" <R.A.Reese@GRI.HULL.AC.UK>
Subject: Re: P value Debate
Content-Type: TEXT/PLAIN; charset=US-ASCII
On Sun, 3 Jun 2001, Mark Casazza wrote:> To assume that a p value of .00049 is somehow "more significant" than a p
> value of .00051 seems to me to be somewhat spurious.
I contributed to the SEMNET debate so won't repeat myself. However, as a
nice example relevant to Mark's point, I looked at a student's results
based on correlations between scales. Personally, I am happy on most
occasions to use Pearson correlation between ordinal scales, but some
examiners get bees in bonnets about parametric versus non-parametric. It
seemed wise for the student to check the values using Spearman's formula.
Unsurprisingly, none of the correlations was substantially changed - the
Pearson and Spearman values differed at the second or third significant
figure. But for one pair of variables, this moved the "significance"
from 0.06 to 0.04. My advice? After Bonferroni adjustment, neither
value was near the usual claim for significance, and other correlations
were significant at roughly 0.001, so needed discussing first.
R. Allan Reese Email: firstname.lastname@example.org
Associate Manager Direct voice: +44 1482 466845
Graduate Research Institute Voice messages: +44 1482 466844
Hull University, Hull HU6 7RX, UK. Fax: +44 1482 466436