Modern Demographic Methods in Epidemiology
with R
Monday 23 November 2015, 9 am - 6 pm.
Melbourne School of Population and Global Health (MSPGH), Australia
Course overview
The course will give an overview of modern
demographic methods in epidemiology, including a practical training in
analysis and reporting study results
with R. The focus will be on
analysis of follow-up studies in continuous time, including handling
of multistate models and competing risks.
Course details
- The intended audience are statisticians and
epidemiologists who want to increase familiarity with basic
and advanced concepts as well as practical tools for modern
epidemiological analysis, primarily in follow-up studies.
- Topics covered:
- Rates, survival function, lifetables:
Connection between basic concepts from a probability point of view.
- Multistate models, competing risks.
- Lexis diagrams and multiple timescales.
- Lexis objects (Epi package) for follow-up, competing
risk and mutistate data
- Cox-modeling of rates.
- Splitting the follow-up and likelihood.
- Poisson modeling of rates.
- Modeling multiple time scales.
- Reporting rates and probabilities from models with
multiple states and timescales.
- The format of the course will be lectures alternating
with practical computer exercises, the latter based on the
free computer
program R,
for which a number of packages for epidemiology are
available.
Students should bring their own laptop with R installed;
see details below under Computing
- Course instructors:
- Lectures & practicals:
Bendix Carstensen,
Senior Statistician, Steno Diabetes Center & Department
of Biostatistics, University of Copenhagen, Denmark.
- Computer practicals:
Lyle Gurrin University of Melbourne, & Bendix Carstensen
- Venue: TBA
- Price: 300 AUD
- Course organizer:
Lyle Gurrin
- Application form is
here; fill it out, scan and mail to
Lyle Gurrin, lgurrin@unimelb.edu.au.
- Course program
- Course material: (some dead links, to appear)
- Computing details
- Formally no prior knowledge
of R is required, a short
introduction is given.
- Participants are assumed to bring their own laptop with the
latest version of R (version 3.2.2) installed; also the following
packages for R should be installed:
- Epi (version 1.1.71)
- mstate
- msm
- How to: While connected to the internet, click on "Packages" select a
mirror (a server site, that is) from where to get the packages
and subsequently select the packages.
- Check the installation by firing up R, typing:
library(Epi);library(msm) and checking the versions of
R and the Epi package by typing sessionInfo()
- If you are not familiar with R, you can use "A work-book
introduction to basic R" by Michael Hills:
First go
here
and click on the link "readme.fst" and follow the instructions.
- A slightly more advanced text is
A short introduction to R for
epidemiology.
- Mailing list:
A mailing list for users of R in epidemiology has been set up.
Join the list and use it for getting help with your R-problems:
Join here.
- Useful papers in pdf-format:
- Previous courses are here.
- VicBiostat are hosting courses on Tuesday 24th through Thursday 26th:
Course background
The fundamental quantities in observation and description in
population processes are the rates of event occurrence. If rates are
known for all types of events of interest (such as occurrence of
marriage, birth, disease or death) it is possible to reconstruct the
entire underlying process and answer relevant questions about rates
and probabilities.
All effects in population processes are inherently continuous even if
we of course only ever see the discrete events (deaths, cancer
diagnoses, marriages) etc. This means that we should use models and
methods that allows approximations of continuous (non-linear) time
effects on rates.
The classical epidemiological approach has been to use 5-year
age-classes, in order to get manageable data and to get results that
are possible to present in a table. This course will show how to use
more disaggregated data to estimate age-effects that are continuous
functions of age. This will be extended to several simultaneous
timescales.
As a consequence of this, the graphical reporting of results is an
important part of the course. In brief, the course will give an
overview of modern demographic methods in epidemiology, including a
practical training in analysis and reporting study results
with R. In particular there will be
focus on estimation and (graphical) reporting of non-linear effects of
time-scales such as age and disease duration.
About the course instructors
-
Bendix Carstensen is a
Senior Statistician in the department of Clinical Epidemiology at
the Steno Diabetes Center in
Copenhagen, and is associated with the department of Biostatistics at
the University of Copenhagen and the department of Medical
Epidemiology and Biostatistics at Karolinska Institutet in Stockholm.
He holds a degree in mathematical statistics from the University of
Copenhagen and has 30 years of experience in practical biostatistics
and epidemiology. He has a keen interest in demographic methods, and is
the maintainer and co-author of
the Epi package
for R, and has over 25 years of
international teaching experience in practical statistics and
epidemiology - of which a lot is available at his website.
- Lyle
Gurrin is Associate Professor of Biostatistics at the School of
Population and Global Health (MSPGH) at the University of
Melbourne. Prior to joining the School in 2003, Lyle spent five years
as a full-time senior biostatistician at King Edward Memorial Hospital
for Women in Perth and the UWA Dept of Obs & Gyn after completing a
PhD in Biostatistics at the Telethon Kids Institute in Perth.
Lyle has extensive experience in applying methods for longitudinal and
correlated data in cohort studies, and runs an on-going
epidemiological study of the genetic and environmental modifiers of
hereditary haemochromatosis (iron overload disease). He promotes sound
epidemiology, emerging biostatistical methods and the principles of
causal inference through teaching biostatistics to postgraduate
coursework students at the MSPGH and the Biostatistics Collaboration
of Australia (BCA), and as a CI of the NHMRC-funded Victorian Centre
for Biostatistics (ViCBiostat).
Last updated 1 October 2015, BxC