MIXED MODELS ANALYSIS OF MEDICAL DATA USING SAS
Edinburgh, 12-14 October 2011
Course content:
This course will cover the statistical background to the mixed model
and will emphasise its practical application in medical data with
particular reference to clinical trials. All analyses will be
illustrated using SAS and lectures will be combined with practical
sessions in order to reinforce concepts. Topics covered include:
Day 1
- General concepts and underlying statistical theory
- Use and interpretation of PROC MIXED
- Multi-centre trials and meta-analysis
- Consideration of issues such as biased standard errors, significance
testing and negative variance components
Day 2
- Repeated measures trials
- Random coefficients models
- Crossover trials
- More complex trial designs
- Introduction to Bayesian methods
Day 3
- Generalised linear mixed models
- Mixed models for ordinal data
- Use and interpretation of PROC GENMOD and PROC GLIMMIX
Note the practical sessions will focus on constructing models and
interpreting results from PROC MIXED output, and will not be "hands-on".
Who should attend:
This course is directed at medical statisticians who wish to
understand the statistical background to mixed models and to carry out
analyses using SAS.
Why attend:
Conventionally, clinical data is analysed using fixed effects models.
However, benefits can often be gained by using a mixed model. For
example: in repeated measures trials full allowance can be made for
the correlation occurring between the repeated observations even if
data are missing; in multicentre trials or meta analyses treatment
standard errors are more appropriately based on between centre/trial
variation (fixed effects standard errors are based on within
centre/trial variation); in crossover trials more accurate treatment
means are often achieved by combining within and between patient
estimates. Suitable procedures are now readily available for fitting
these models well known packages such as SAS. This has led widespread
application and knowledge of mixed models becoming essential for
medical statisticians.
As with any statistical technique a firm understanding of the
theoretical background is essential to allow its effective application
and to obtain a clear interpretation of results.
Course fees:
Standard rate £850
Registered charities/academic institutions £650
Fees include lunches, a course dinner and a delegate's pack including
full course notes.
The speakers:
Helen Brown is a Senior Statistician at the University of Edinburgh
and has a research interest in the use of mixed models in medicine.
She has over twenty five years of practical experience as a
statistician and has been employed within academia, the health service
and the pharmaceutical industry. She is co-author of Applied Mixed
Models in Medicine, in the John Wiley Statistics in Practice series.
Eleanor Allan is a Principal Statistician in the Statistical Services
Centre (SSC) at the University of Reading. She has been a consulting
statistician for over 25 years, 10 of which were spent in the
pharmaceutical industry. Her main statistical interests include mixed
modelling, multilevel modelling, repeated measurements analysis and
the design and analysis of experiments.
Venue:
The course will take place in the Holiday Inn, Edinburgh-North, one
mile from the city centre and easily accessible from the main railway
station and airport.
Accommodation:
Course participants have the opportunity to stay at the Holiday Inn,
Edinburgh-North at a reduced rate. All bedrooms in the hotel have
modem points.
For further information please contact the Reservations department on
0131 332 2442 quoting University of Edinburgh. Alternatively please
contact our accommodation agency: Murray Accommodation, telephone
08707 509808, or visit www.murray-accommodation.co.uk.
Further information:
Course website: http://www.lifelong.ed.ac.uk/mixedmodels/
Course administrator: [log in to unmask] or telephone +44 (131) 651 1821.
--
The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.
You may leave the list at any time by sending the command
SIGNOFF allstat
to [log in to unmask], leaving the subject line blank.
|