[Rd] historical significance of Pr(>Chisq) < 2.2e-16

peter dalgaard pdalgd at gmail.com
Thu May 8 09:35:27 CEST 2014

On 08 May 2014, at 08:38 , Prof Brian Ripley <ripley at stats.ox.ac.uk> wrote:

> Only in very unusual circumstances does it matter how small a p-value less than, say, 1e-6 is.

Slight disagreement: It does matter when there are multiple comparisons in the order of several thousands, which is not so unusual in bioinformatics at least. The better scientists in the field seems to know better than to trust extreme-tail behaviour of test statistics, though. 

(In the case I'm thinking of, they had a two-group comparison for several thousand DNA markers. One might Bonferroni-correct and multiply the individual p-values by the number of comparisons, but they got the rather better idea of calculating an overall test statistic as the maximum over all marker points and then finding its permutation distribution.)

Peter Dalgaard, Professor,
Center for Statistics, Copenhagen Business School
Solbjerg Plads 3, 2000 Frederiksberg, Denmark
Phone: (+45)38153501
Email: pd.mes at cbs.dk  Priv: PDalgd at gmail.com

More information about the R-devel mailing list