Dear all,
1. On Wednesday 26th January, the RSS Leeds/Bradford local group will be
hosting an afternoon of talks on "Statistics and data linkage" featuring
Thomas Fleming (Centre for Epidemiology and Biostatistics, University of
Leeds), Gemma Catney (Centre for Public Health, Queens University
Belfast) and Myles Gould (School of Geography, University of Leeds).
The meeting will be held in the Leeds University Worsley Building room
9.57 from 2pm to 5pm with refreshments available from 1.30pm.
2. On Monday 14th February, the RSS Leeds/Bradford local group will be
hosting a talk on "Regression models in space and time" by Adrian Bowman
(School of Mathematics and Statistics, University of Glasgow).
The meeting will be held in the Leeds University Roger Stevens Building
room 14 from 4pm to 5pm with refreshments available from 3.30pm in the
level 9 foyer of the School of Mathematics.
Further details of both events can be found on our webpage:
http://tinyurl.com/rss-lba
Regards, Paul
===================================================================
Dr. Paul D. Baxter
Secretary/Treasurer, RSS Leeds/Bradford Local Group,
Division of Biostatistics, University of Leeds, Leeds, LS2 9JT, UK.
-------------------------------------------------------------------
Leeds/Bradford: Wednesday 26 January, 2.00pm, University of Leeds.
Statistics and data linkage
Thomas Fleming (University of Leeds)
Record linkage: an overview of the two main techniques and a comparison
of their use in Hospital Episodes Statistics data
Record linkage is the technique of taking two datasets and determining
which entities in one dataset relate to those in the other. Also known
as data matching, it is often used to extend the number of variables in
the first dataset by linking to the second dataset or to fill in gaps in
variables in the first dataset that also exist in the second dataset.
In this talk I will present an overview of the two main classes of
record linkage techniques, deterministic and probabilistic.
Deterministic techniques start with defining a list of matching rules
and iterating through to decide if two records match. Probabilistic
techniques take a much more statistical approach by comparing two
records and calculating a match weight.
I will then move on to a comparison of the two techniques using Hospital
Episodes Statistics (HES). HES is a dataset containing information on a
patient's episode of care whilst in hospital. The unique patient
identifier HESID is assigned using a deterministic record linkage
algorithm, in essence linking together all hospital episodes for a
patient. I will describe the rules set for this and compare it to the
use of a probabilistic record linkage algorithm. This will be
illustrated in an extract from HES containing 400,000 episodes.
Knowing the characteristics of how record linkage has been done is
important for any analysis on a linked dataset so decisions can be made
with the full knowledge when modelling.
Gemma Catney (Queens University Belfast)
The Northern Ireland Longitudinal Study: data linkage, research
potential and application
This seminar is divided into two parts; the first provides background to
two major data linkage studies, the Northern Ireland Longitudinal Study
(NILS) and Northern Ireland Mortality Study (NIMS). The NILS is a
large-scale, representative study created by linking the Northern
Ireland Health Card Registration system to 2001 Census returns and
administrative data from other sources. These include vital events
registered with the General Register Office for Northern Ireland (i.e.,
births, deaths and marriages) and the Health Card registration system
migration events data. Selection into the NILS is based on birth date
and sample is large - c.28% of the Northern Ireland population
(approximately 500,000 individuals). NIMS is another large-scale data
linkage study linking 2001 Census returns for the whole enumerated
population (approximately 1.6 million individuals) to subsequently
registered mortality data from the General Register Office (GRO-NI).
The second part of the talk is an application of NIMS data to explore
the relationship between population clustering by religion and health.
Little is known about the effect of the social environment on health in
Northern Ireland, where there is marked segregation and a large
proportion of the population remain clustered residentially into
Catholic or Protestant majority areas. Using the self-reported health
measure in the 2001 Census of Population and longitudinal data on
mortality risk in the six years following the Census the analysis
considers if those who reside in an area in which they are in a
‘religious’ majority fair better in health terms. Key questions include:
is there a positive effect of living with one’s ‘own’ group, where there
may be greater intra-group social capital and better developed social
support networks? Or does residing in a segregated area have a negative
impact on health, whereby unhealthy behaviour and a culture of
ill-health may perpetuate? Are these effects the same for Catholic- and
Protestant-dominated areas, and what is the effect of population mixing?
Controlling for socio-economic confounders at the individual and area
level, we address these questions using a measure of population
concentration (% group) and segregation indices.
Myles Gould (University of Leeds)
Modelling short-distance residential moves using linked data: An
application of the Northern Ireland Longitudinal Study (NILS)
This presentation uses the Northern Ireland Longitudinal Study (NILS) -
comprising data on Health Card registrations linked to the 2001 Census -
to analyse residential migratory moves in Northern Ireland between 2001
and 2007 (for 28% of this population). The analysis considers individual
and neighbourhood factors that shape whether individuals change
residential address or not, and also the distances moved. Examples of
questions to which answers are sought include whether there are
differences in mobility by religion, by marital status, limiting
long-term illness status; and whether individuals in areas where they
are in `the minority` (e.g. Catholics in Protestant areas) are more
mobile than those in places where they are `the majority`. Multilevel
modelling is used to explore the determinants of both the probability of
moving and the distance moved; the relative importance of between
individual and between-place variability, and also the interactions
between place and Individual characteristics. The main findings are that
patterns of movement are strongly structured by religious denomination
and by socio-economic background, and that `place` (in terms of Super
Output Area of residence) is an important context. Other results also
indicate that public authority housing residents are less mobile than
others; that there are differentials in mobility by age, education and
Limiting long-term illness; and that migration does not have a major
impact on redistributing population with regard to community background.
The meeting will be held at Leeds University Worsley Building on Level
9, Room Worsley SR (9.57), starting at 2pm with refreshments from 1.30pm.
-------------------------------------------------------------------
Leeds/Bradford: Wednesday 14 February, 4.00pm, University of Leeds.
Regression models in space and time
Adrian Bowman (University of Glasgow)
Additive, and more general nonparametric, approaches to modelling extend
standard regression methods by allowing very flexible, but smooth,
relationships between variables of interest. These models are
particularly helpful in environmental applications, where there is a
need to model complex forms of spatial and temporal trends, as well as
spatial and temporal correlation. Technical aspects of the talk will
include computational strategies for spatiotemporal smoothing and ways
of extending standard inferential methods. The data structures
considered will include river networks as well as more standard spatial
domains. Applications will include the modelling of SO2 pollution over
Europe, water quality in the River Tweed and rainfall-flow response in
the river Dee.
The meeting will be held at Leeds University Roger Stevens Building in
room RSLT14 at 4pm with refreshments from 3.30pm in the level 9 foyer of
the School of Mathematics.
You may leave the list at any time by sending the command
SIGNOFF allstat
to [log in to unmask], leaving the subject line blank.
|