GENEALOGY-DNA-L Archives

Archiver > GENEALOGY-DNA > 2010-11 > 1290243233


From: "Lancaster-Boon" <>
Subject: [DNA] Odds Are, It's Wrong - 5% of the time
Date: Sat, 20 Nov 2010 09:53:53 +0100


Hi Steve

I guess that there is a legitimate point to be made, and it is connected to
the point Ken now made, that in our time there are now an enormous number of
people who have learnt how to manipulate numbers and formulas and get those
manipulations through peer review, but it is possible to be skilled in doing
this without being skilled in interpretation back into the real world.

Very often reports about tolerance intervals etc are almost meaningless
because they compare to a model of reality that is itself not certainly
correct. That is of course not how it is when we talk about rolling a dice
or flipping a coin, because the "models" of the probabilities on each roll
or flip are simple and well-known.

So, in summary, flipping a coin or rolling a dice, while useful for learning
the maths, is not so perfect for explaining the difficulties of
interpretation of the maths back into real world models.

I hope what I have written makes some level of sense! :D

Best Regards
Andrew


===========
From: Steven Bird <>
Subject: [DNA] Odds Are, It's Wrong - 5% of the time
Date: Fri, 19 Nov 2010 20:00:17 -0600
References: <>,
,<>,<COL118-W33D30F8531A4233CE0ED
>,<>
In-Reply-To: <>

The cumulative odds of losing three times in a row, given a 19 out of 20
chance to win on each play, are exactly what I stated. .05*.05*.05=.000125
or .00205% The odds of each chance remain 19 out of 20, but cumulative odds
of winning ONE time increase. If I flip a fair coin, the chance of heads is
50/50. However, the chance of NO heads in ten tosses diminishes to 0.5^10,
or .000977.

As I indicated before, if you are unhappy with a 95% chance of being right,
then change the p value to 99.7% or 99.9999%. It is strictly up to the
statistician. There is nothing sacred about p=0.05

You might be right about the double or nothing bet however. ;-)



This thread: