You need to use generalized linear models. see McCullagh & Nelder, Generalized linear models. Chapman and Hall
-----Original Message-----
From: A UK-based worldwide e-mail broadcast system mailing list on behalf of Automatic digest processor
Sent: Sat 11/09/2004 00:00
To: Recipients of allstat digests
Cc:
Subject: allstat Digest - 9 Sep 2004 to 10 Sep 2004 (#2004-209)
There are 8 messages totalling 352 lines in this issue.
Topics of the day:
1. Query: sample size for dose finding study (2)
2. University of New South Wales Postdoctoral Fellowships
3. JOB: SAS Programmer
4. Oxford Vacancy: Dept of Statistics in assoc. with St John's
College-Lectureship in Bioinformatics: correction
5. Weighted Regression to fix non-constant variance
6. BIOSTATISTICIANS WANTED
7. Query: computing standard errors of partitioned data
----------------------------------------------------------------------
Date: Thu, 9 Sep 2004 11:10:34 +0100
From: Michael Hutchinson <[log in to unmask]>
Subject: Query: sample size for dose finding study
I was hoping someone on the list could provide reference(s) re estimating=
=20
sample size when the objective is estimation of the maximum tolerated dos=
e=20
(dose escalation is within patients). I ran a search on medline but didn=92=
t=20
find anything useful (the most relevant paper found didn't describe or=20
justify the sample size calculation) and Machin, Campbell et al. (where I=
=20
usually look first) doesn't cover it. Any info would be much appreciated.
Thanks.
Mike
_________________________________________________________________
Express yourself with cool new emoticons http://www.msn.co.uk/specials/my=
emo
------------------------------
Date: Thu, 9 Sep 2004 16:01:09 +0100
From: Michael Hutchinson <[log in to unmask]>
Subject: Query: sample size for dose finding study
I was hoping someone could provide reference(s) re sample size calculations
for a dose finding study when the objective is estimation of the maximum
tolerated dose (dose escalation is within patients). I've run a search on
medline but it didn't reveal anything useful.
Any info would be much appreciated.
Thanks.
Mike
_________________________________________________________________
Use MSN Messenger to send music and pics to your friends
http://www.msn.co.uk/messenger
------------------------------
Date: Fri, 10 Sep 2004 17:15:19 +1000
From: Robert Kohn <[log in to unmask]>
Subject: University of New South Wales Postdoctoral Fellowships
Postdoctoral Fellowships at the University
of New South Wales
The University of New South Wales is offering Postdoctoral
Fellowships available from January 2005 for a period of 3 years.
Professors Robert Kohn and Michael Sherris and their groups in
the Faculty of Commerce and Economics would like to encourage
applications focusing on Multivariate Modeling and Actuarial
Risk Management using Copula Models.
Applicants must hold a doctorate at the time of application
(October 1, 2004)and must not have been awarded their
doctorates more than three years ago. Applications close on
October 1, 2004.
For details of the positions see
http://www.ro.unsw.edu.au/funding/vcpdf05.shtml
and below.
Successful candidates in this field will automatically
be eligible to receive the Faculty of Commerce and
Economics post-doctoral fellowship supplement of
$10,000 pa available to recipients of a NewSouth Global
or Vice-Chancellor's Post-Doctoral Fellowship in the
Faculty of Commerce and Economics. No separate application
is required. Applicants should apply for the NewSouth Global
and Vice-Chancellors Post-Doctoral Fellowships.
We would be grateful if you would pass this on to anyone
that may be interested in applying for one of these
Post-Doctoral Fellowships to work at the Faculty of
Commerce and Economics on the following project:
Multivariate Modeling and Actuarial Risk Management
using Copula Models.
This work will be carried out with one or more the following:
Professor Robert Kohn ([log in to unmask])
Professor Michael Sherris ([log in to unmask])
Professor Emil Valdez ([log in to unmask])
Contact one of the above for further details.
Essential criteria are an outstanding academic background,
publications in international journals, a strong
background in both Mathematics and Statistics,
programming skills in Matlab or a related language.
Desirable criteria are knowledge of Bayesian statistics
and/or Financial and Insurance Risk Modelling including
Actuarial Science.
Professor Robert Kohn
Faculty of Commerce and Economics
School of Economics
John Goodsell Building
Room 222
University of New South Wales
UNSW Sydney 2052 NSW
Australia
Phone 612 9385 2150
Fax: + 612 9313 6337
Email [log in to unmask]
------------------------------
Date: Fri, 10 Sep 2004 12:49:32 +0100
From: George Vernon <[log in to unmask]>
Subject: JOB: SAS Programmer
SAS Programmer
Salary: Very Competitive =09
=09
Business Type: Biopharmaceutical Company, Biotechnology Company,
Clinical Research Organisation, Contract Research Organisation,
Pharmaceutical Company =09
=09
Position type: Permanent =09
Location: Ireland =09
=09
Background: =09
=09
My client has two open positions for SAS Programmers. Candidates will
work within a mid-sized biostatistics team operating closely alongside
senior statisticians. Candidates must have 2+ years working within a
similar position in a pharma/CRO. =09
=09
Additional Information: =09
=09
Hobson Prior is a corporate member of the REC and operates strictly
within the regulations governing the conduct of employment businesses.
This requires us to provide detailed information to candidates in
relation to specific roles prior to the submission of their personal
details and prohibits the disclosure of information relating to
candidates without their consent =09
=09
To apply, please attached an up to date copy of your CV quoting the
reference 'grv-1893'. =09
=20
George Vernon
Hobson Prior
=20
t: +44 1892 612612
f: +44 1892 612613
e: [log in to unmask]
<mailto:[log in to unmask]>=20
=20
This message is for the intended recipient only. It may contain
confidential or proprietary information. If you receive this message in
error, please immediately delete it, destroy all copies and notify the
sender. You must not use or disclose any part of this message if you are
not the intended recipient. We may monitor all email communication
through our networks. Any views expressed in this message are those of
the individual sender, except where the message states otherwise. We
take reasonable precautions to ensure our emails are virus free.
However, we cannot accept responsibility for any virus transmitted by us
and recommend that you subject any incoming email to your own virus
checking procedures.
=20
=20
=20
=20
------------------------------
Date: Fri, 10 Sep 2004 13:20:03 +0100
From: Anna Beint <[log in to unmask]>
Subject: Oxford Vacancy: Dept of Statistics in assoc. with St John's College-Lectureship in Bioinformatics: correction
The closing date for this vacancy has been corrected to 22 October 2004
UNIVERSITY OF OXFORD
Lecturership in Bioinformatics
Department of Statistics in association with St John's College
The Department of Statistics is experiencing an exciting period of=20
growth and development, and is one of the leading UK departments.
The University seeks to appoint a lecturer in bioinformatics to take up=20
this post from 1 January 2005, or from a mutually agreed later date. The =
Lecturership will be held in conjunction with a Supernumerary Fellowship =
at St John's College.
The University interprets bioinformatics widely, to include the whole=20
range of applications of mathematical, statistical, or computational=20
techniques to the analysis of data arising in modern molecular genetics. =
Whilst applications are welcomed from candidates with research interests =
in any area of methodological development in bioinformatics, there is a=20
strong preference for candidates working on post-genomic problems which=20
complement or build on existing strengths in bioinformatics within the=20
Department of Statistics. These include structural bioinformatics,=20
comparative genomics, statistical alignment, population genomics and=20
genetic variation, human disease studies, and analytical tools for=20
modern experimental techniques such as gene expression arrays,=20
proteomics, metabonomics etc.
The University salary for the post is on a scale up to =A345,707 p.a.=20
Additional college allowances are available as set out in the further=20
particulars. This post is in an area currently designated as a shortage=20
subject under the HEFCE "Golden Hello" scheme. Appointees may therefore, =
under certain conditions, be eligible for a 3-year salary supplement.
Entitlement to sabbatical leave accrues at the rate of one term's leave=20
for every six terms with normal duties.
Further particulars can be found in .pdf format at the link below or=20
contact the Personnel Administrator, Department of Statistics, 1 South=20
Parks Road, Oxford OX1 3TG (Tel 01865 272860), email=20
[log in to unmask] The closing date for applications is Friday 22th=20
October 2004.
The University of Oxford is an Equal Opportunities Employer
http://www.stats.ox.ac.uk/jobs/StJohnfp.pdf
------------------------------
Date: Fri, 10 Sep 2004 11:43:09 -0400
From: Regina Malina <[log in to unmask]>
Subject: Weighted Regression to fix non-constant variance
Hi everyone,
I am building a linear regression model and I found that, when I plot the
residuals (y axis) versus the predicted values (x axis), I get something
that appears linear with the variance increasing as the x value increases
(wedge shape). I have found the following recommendation for fixing this
kind of problem of non-constant variance (following is my interpretation of
it):
run unweighted regression and save predicted values and residuals,
calculate the variance of residuals at each predicted value,
calculate reciprocal of the variance (this is the Weight),
merge this Weight back to the original dataset via predicted value,
run weighted regression (using the Weight variable calculated),
resulting residual plot should have more constant variance across
predictor point.
Does anyone have experience with this approach? Did I understand this right?
Any other ideas to fix my problem (I have already tried several
transformations of predictors)?
Thank you in advance!!!! Regina
------------------------------
Date: Fri, 10 Sep 2004 17:08:51 +0100
From: Bryan Mackie <[log in to unmask]>
Subject: BIOSTATISTICIANS WANTED
NEW VACANCY +NOW IS THE TIME TO MAKE THE MOVE +NEW VACANCY
THE LOOT: SALARY FROM =A335 K + EXELLENT COMPANY BENEFITS
THE LOCATION: BASED IN THE SOUTH EAST.
THE LIFESTYLE:IF YOU ARE LOOKING TO DEVELOP YOUR CAREER WITH A GLOBAL CRO
WHAT ARE YOU WAITING FOR ,THIS IS IT!
THE LOWDOWN: HAVE 3+ YEARS EXPERIENCE ,
PLEASE E MAIL ME WITH YOUR UPDATED CV ,
I LOOK FORWARD TO HEARING FROM YOU,
BRYAN,
[log in to unmask]
------------------------------
Date: Fri, 10 Sep 2004 09:44:32 -0700
From: richard bowyer <[log in to unmask]>
Subject: Query: computing standard errors of partitioned data
Hi All,
I would like to prove the followin situation algebraically:
I have a random sample X1, X2, X3, ...., XN with grand mean M, standard
deviation S and standard error SE_N=S/sqrt(N)
I divide my sample into P partitions and compute the means of each partition
m1, m2, ..., mP. Define M'=mean(m1, m2, ..., mP), the mean of the means,
which is simply the grand mean i.e. M'=M. Let S_P be the standard deviation
of these P partition means and SE_P=S_P/sqrt(P) the standard error (of M').
Clearly As P-> N SE_P -> SE_N in other words
As P-> N S_P/sqrt(P) -> S/sqrt(N) [1]
I would like to demonstrate [1] algebraically and also determine the rate of
convergence
Many thanks in advance for your help
regards,
Richard.
_________________________________________________________________
Express yourself instantly with MSN Messenger! Download today - it's FREE!
http://messenger.msn.click-url.com/go/onm00200471ave/direct/01/
------------------------------
End of allstat Digest - 9 Sep 2004 to 10 Sep 2004 (#2004-209)
*************************************************************
|