Short course
by Michael Höhle,
Department of Mathematics, Stockholm University, Sweden.
Dates
18-19 June 2014, Each day
from 09:00-16:00.
Venue
Robert Koch Institute,
Seestraße 10, Berlin, Germany. Room 2.037.
Course motivation
Public health authorities
have, in an attempt to meet the threats of infectious diseases,
created comprehensive mechanisms for the collection of data on
cases of infectious diseases. The vast amounts of acquired data
demand appropriate statistical methods describing the dynamics and
the development of algorithms for the automated detection of
abnormalities.
This short course covers statistical aspects of how to model and
monitor routine collected surveillance data which, depending on
the temporal and spatial scale, can be seen as realizations of the
following stochastic processes:
- discrete space -
discrete time process, i.e. univariate or multivariate count
data time series
- discrete space -
continuous time point processes
- continuous space -
continuous time point processes
With a strong emphasis on
the simplest structure - the univariate count data time series -
the course presents methods for the
retrospective and prospective analysis. An implementation aspect
of the methods is given by applications using the R package
'surveillance'.
Course contents
- Motivating
examples: Why is there an interest in the modelling and
monitoring of routinely collected public health data.
- Overview of
the R package 'surveillance'
- Looking at
what is there: in univariate and multivariate count data time
series: trends, change-points etc.
- Looking at
what is not there: latency periods and reporting delays
-
Back-calculation method
-
Nowcasting
- Prospective
aberration detection: The algorithm of Farrington (1996) and
beyond.
-
Endemic-epidemic two component modelling: Three views in time
and space-time
The course content will
be illuminated both from a theoretical and an applied RKI
perspective. In order to enhance the practical understanding of
the methods, R code is given where possible - especially, the R
package "surveillance" will be used. The course format encourages
active discussions on how to apply the presented modelling
techniques in an RKI perspective.
Lecture material
Course slides (2x2 version) - Note: The
page is password protected.
Links
