Archiver > GENEALOGY-DNA > 2006-01 > 1138230821

From: "Sasson Margaliot" <>
Subject: Re: [DNA] Ashkenazi levites Y modal haplotype (L MH)
Date: Thu, 26 Jan 2006 01:13:41 +0200
References: <BAY7-F7FB3DFCFC27B203237CE5B4130@phx.gbl> <000501c620f3$b2c6c2b0$79bf19ac@sasonb46c858c2> <> <003e01c62146$b9621cc0$79bf19ac@sasonb46c858c2> <>

John Chandler wrote:

> If, as you suggest, the second result is actually
> outside of the 95% confidence interval for the first, then that very
> likely constitutes a significant difference between the two, and the
> overlap is not at all comforting.

I do not know what was actual 95% confidence interval. I have just used
arbitrary numbers to illustrate the effect of "double standard deviation"
for those readers of the List who are not familiar with this concept. With
the sample size of just 80 Levites the "double standard deviation" range is
probably much wider than examples I gave.

> A signifcant difference on the
> *same* *experiment* means there was a serious error somewhere --

The "apparatus" (in extended sense) available and used in 1998 is not at all
the same as in 2003, so it is not the same experiment

> either in including those "questionable" data in the first place or in
> throwing them away for the second result.

The first set of data were according to the state-of-art level of 1998, at
which time they were not "questionable" at all.

The second set is according to the standards of 2003, and not all records
were accepted according to reasonable criteria appropriate at 2003.

In any case, in addition to the "regular" (statistical) standard deviation,
there is also difficulty of extracting *representative sample*, so that the
observed difference is as expected and does not imply any error in handling
the data.

Sasson Margaliot

This thread: