HOME

Sample Discussion

Subject: P-value and Type I error

From drusarkar_at_rediffmail.com  Thu Jul 14 09:57:44 2005
Date: 14 Jul 2005 10:16:47 -0000
From: "uttam sarkar" <drusarkar_at_rediffmail.com>
Subject: "Help"
To: Multiple Recipients of <angenmap_at_animalgenome.org>

Dear Sir,

My question is:
How standard deviation (SD) of gene frequency is estimated for a particular experiment?

For example in GH gene
Breed     N     A allele   B allele    SD
Nellore   63    0.91       0.09        ?

Thank you.
Dr. Uttam Sarkar
Ph. D. scholar (AGB)
NDRI, India
From chang_at_calshp.cals.wisc.edu  Thu Jul 14 10:15:59 2005
Date: Thu, 14 Jul 2005 10:15:59 -0500 (CDT)
From: ruby chang <chang_at_calshp.cals.wisc.edu>
Subject: type I error
To: Multiple Recipients of <angenmap_at_animalgenome.org>

Hi,

Type I error (alpha) is usually set at 0.05 level. I find it hard
to convince some people that an effect can still be "significant" with a
p-value of 0.06.

Does anyone know when and where this magic number 0.05 started?

Thanks.

Ruby Chang
Dairy Science Department
University of Wisconsin, Madison
From DNADave_at_U.Arizona.Edu  Thu Jul 14 10:28:59 2005
Delivered-To: angenmap_at_genome.ansci.iastate.edu
Subject: Re: type I error
From: David Henderson <DNADave_at_U.Arizona.Edu>
Organization: The University of Arizona
Date: Thu, 14 Jul 2005 08:28:35 -0700
To: Multiple Recipients of <angenmap_at_animalgenome.org>


> Type I error (alpha) is usually set at 0.05 level. I find it hard
> to convince some people that an effect can still be "significant" with a
> p-value of 0.06.

You are correct.  If you take a Neyman-Pearson view of statistics and
you want to control your Type I error rate at 5% (for a single test),
then you reject only when p < 0.05.  If you take the view that Fisher
intended, then p-values are only measures of risk whereby p = 0.06 is
slightly more risky than p = 0.05, but both could equally be significant
in the eye of the right beholder...

> Does anyone know when and where this magic number 0.05 started?

R.A. Fisher somewhere around 1920.  A good read on this is "The Lady
Tasting Tea" which describes the story around 0.05 becoming a standard.
A famous quote by Fisher is "Surely God loves the 0.06 as much as the
0.05.".

Good Luck!!

Dave H
-- 
David A. Henderson, Ph.D.
Insightful Corporation
1700 Westlake Avenue North, Suite 500
Seattle, WA 98109-3044
Tel: 206-802-2307
Fax: 206-283-8691
DNADave_at_Insightful.Com
From david.machugh_at_ucd.ie  Thu Jul 14 11:23:53 2005
Date: Thu, 14 Jul 2005 17:05:38 +0100
From: David MacHugh <david.machugh_at_ucd.ie>
Subject: RE: "Help"
To: Multiple Recipients of <angenmap_at_animalgenome.org>

Hello Uttam,

The variance of an allele frequency estimate is (from the binomial
distribution):

p(1-p)/2n

Where n = the number of animals sampled (half the number of chromosomes
assayed for a diploid marker).

The SD can be calculated as the square root of this.

In your case the variance for both allele frequency estimates would be:

0.91(0.09)/126 = 0.00065

The SD would be then be the square root of 0.00065 = 0.0255

Note, the formula is exactly the same if you have more than two alleles (as
is the case for most microsatellites).

Hope this helps,

David.

David MacHugh, Ph.D.
SFI Investigator,
Animal Genomics Laboratory,
School of Agriculture, Food Science and Veterinary Medicine,
and Conway Institute for Biomolecular and Biomedical Research,
College of Life Sciences,
University College Dublin,
Belfield, Dublin 4.
Ireland.

Phone: 353-1-716-7738
Mobile: 353-87-202-9763
Fax: 353-1-716-1103
Email: david.machugh_at_ucd.ie
Website: http://www.ucd.ie/ascience/html/dmachugh-frame.html
Conway website: http://www.ucd.ie/~conway/

Go back to the AnGenMap main page.

 

 

© 2003-2024: USA · USDA · NRPSP8 · Program to Accelerate Animal Genomics Applications. Contact: Bioinformatics Team