ROYAL STATISTICAL SOCIETY : MEDICAL SECTION
Tuesday 30 September 2003, 4.00 at the RSS (Tea at 3.30)
To be held at the RSS, 12 Errol Street, London EC1Y 8LX
(map http://www.rss.org.uk/about/direction.html)
2 talks on .......
'WORKING AS A STATISTICIAN IN A DEVELOPING COUNTRY'
'A nonparametric mixture model for stratified survival data, with an
application to the study of first marital unions in Malawi'
SAMUEL MANDA (University of Leeds)
'Broad but not deep - stretching statistics to meet needs in Nepal'
ALISON ANDERSON (International Nepal Fellowship)
Synopses
In sub-Saharan Africa, age at first marriage is an important explanatory
factor linked to the region's fertility transition. Up-to- date
information on these substantive issues is provided from the Demographic
and Health Survey (DHS) program. The DHS uses a stratified sampling
design to obtain the required sample. For time to event data such as age
at first marriage, a stratified proportional hazards regression is
commonly employed. Sometimes, stratum- specific dummy variables are
included in the regression analyses.
On the other hand, frailty models, which include independently and
identically distributed stratum-specific random variables, can be
thought of as a compromise between stratified and unstratified analyses.
In this talk, we will be presenting a different approach, which treats
the whole stratum-specific baseline hazard function as a random variable
drawn from a population of hazard functions, thus
providing infinite alternatives. We model the whole baseline hazard
function nonparametrically using a mixture of triangular distributions.
The unknown number of mixands and the other parameters are estimated
using a Bayesian approach via the
reversible jump Markov chain Monte Carlo (MCMC) algorithm. Data on
transition to marriage in Malawi is used to illustrate the proposed
methodology. The results are compared to those obtained under the
alternative beta mixture model and parametric
assumptions on the baseline hazard function.
RELEASE is one project of the International Nepal Fellowship, focusing
on clinical and social rehabilitation of people affected by physical
disability, leprosy or HIV/AIDS, predominantly among the poor and
marginalised of the Western Region of Nepal. Appropriate targeting of
resource to maximise real benefits requires good management information,
but management information is itself a scarce resource. Using simplified
statistical methods to help staff to 'turn data into information' has
produced tangible benefits and significantly affected the planning of
the project.
In leprosy control, it is possible to misinterpret failure of the health
system as the desired outcome of disease elimination. Teaching staff to
measure and interpret several indicators together, in order to produce a
composite verbal picture of the situation for use by management, has
allowed the programme to focus on locations with a significant disease
burden.
Prevention, rather than treatment, of nerve damage in leprosy would
reduce the burden of disability and the need for rehabilitation.
Simplification of a double blind clinical trial until it could be run at
a field level and analysed with the resource available, enabled testing
of a hypothesis regarding prophylaxis, and highlighted sub populations
at increased risk of nerve damage.
In absence of valid measures of the impact, programmes can become
activity centred. Development of a culturally appropriate scale to
measure 'participation' allowed programmes to assess the impact of the
rehabilitation process from the client's perspective and revise the
activities appropriately.
Making a variety of statistical methods available to the health care
professionals, through simplification, teaching, training and
consultancy has encouraged staff to look for ways to improve the
technical validity of what they do, with a concomitant increase in focus
and effectiveness of the project.
--
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.
Mailscanner thanks transtec Computers for their support.
|