(this notice is also available in word format)
THE CAMBRIDGE STATISTICS DISCUSSION GROUP
Tuesday 3rd October 2006 7:15 for 7:45
Large Lecture Theatre,
Cognition and Brain Sciences Unit,
Chaucer Road, Cambridge
BLUP combines effect size and significance
Nick Galwey
GlaxoSmithKline, Harlow
Abstract:
Analysis of the association between a disease phenotype and a genetic
marker locus typically gives an estimate of the magnitude of the association
effect (e.g. an odds ratio or a regression coefficient) and a measure of
its significance (a chi-square or t statistic, with its associated p value).
Both of these statistics are of importance when deciding which loci are
worth investigating further. A large estimated effect is of little
interest if the estimate has low precision and a high p value: the effect
may well not be real. Conversely, even a highly significant effect is of
little interest if it is very small. It would be of value to combine the
estimated effect and the significance level into a single statistic: such a
statistic is provided by the Best Linear Unbiased Predictor (BLUP) or
shrunk estimate. The concept of the BLUP, which is supported by a
well-established body of statistical theory, will be illustrated by
considering the effect of a genetic marker locus on a continuous phenotypic
trait. The BLUP will then be used to evaluate the association between each
of a set of genetic markers (SNPs) and a quantitative trait in a sample of
individuals. The ranking of the SNPs as candidates for further
investigation on the basis of the BLUP is shown to be closely related to,
but not identical to, that obtained from the regression coefficient or the
p value. It appears to be straightforward to extend the concept of
the BLUP to the analysis of categorical phenotypes (e.g. to odds ratios
in a case-control study).
Speaker:
Nick was a demonstrator and lecturer at the University of Cambridge
between 1979 and 1995. He then moved to Perth to lecture at the University
of Western Australia from 1995 to 2001 before working as a statistical
geneticist for Oxagen Limited for two years. He has worked in a similar
capacity for GlaxoSmithKline since 2003.
Annual General Meeting: Nick's talk will be preceded by a brief
Annual General Meeting.
Directions: Chaucer Road is off Trumpington Road - if you are heading
away from the city centre it is the first right after the junction
with Brooklands Avenue. Number 15 is at the end on the left and is a
large Victorian House with a more modern extension to its right.
The entrance is in the porch located in the middle of the older section on
the left as you come in and is accessed by ascending a few steps from the
car park. Cars may be parked there. The outside door will be locked at 7:45.
Arrivals after 7:45 may gain admittance admittance by contacting the
secretary on 07761769436.
Next Meetings:
29th November - Sheila Bird (Biostatistics) on ''.
6th February - Julia Gog (Zoology).
7th March - Zoubin Ghahramani (Engineering).
4th April - Matt Whiley (Amgen).
Supper: Some members eat regularly in the University Centre before
each talk meeting in the downstairs bar at 5-45pm. Feel free to join them.
Secretary: Peter Watson, MRC Cognition and Brain Sciences Unit,
15 Chaucer Road, Cambridge CB2 2EF;telephone 01223 355294 Extension 801;
E-mail [log in to unmask]
FOR SECURITY PURPOSES COULD ANYONE INTENDING TO ATTEND THE MEETING PLEASE
LET PETER WATSON KNOW.
|