Short Courses in Applied Statistics

Course outline

This course will introduce capture-recapture techniques and models, and highlight their importance in the application fields of medicine and public health, and the life and social sciences. Such methods aim to address questions such as:
How complete is a register?
How many people have Type II diabetes (diagnosed and undiagnosed)?
How many people possess an illegal fire-arm?
How many illegal drug-users are there in a city?
How many medical professionals are alcohol-dependent?

Historically the purpose of capture-recapture (or mark-and-recapture) procedures was to determine the size of an unknown animal population. However, the methods may be applied more widely and may be used to estimate the size of a human population with a certain disease or a population which is difficult to approach such as a population involved in an illegal activity.

Day 1 of the course introduces the classical Lincoln-Petersen Estimation technique for the situation of two information sources such as hospitals or laboratories, if the characteristic of interest is a disease. Typically, registers, or more generally lists, are available and these are combined to provide a population size estimate. Extensions to more than two sources, leading to log-linear modelling, will be explained. Practical problems will also be discussed, along with the advantages of more than two sources. Practical work will be based around the use of Stata. The CARE software will also be considered.

Day 2 of the course introduces capture-recapture techniques in continuous time. Often recaptures arise not at fixed points in time but rather within a given time window. Typically, one source of information is used to repeatedly count the same individual, such as drug rehabilitation centres which record voluntary patient visits over time. Alternatively, when interest focuses on a criminal population, police data will allow the number of times an individual has been arrested to be counted. The course will explain how capture-recapture methodology can be applied in this context. In particular, a number of approaches which incorporate unobserved heterogeneity will be discussed. Practical work will make use of the SPADE software.

Who should attend?

The course is aimed at statisticians, biostatisticians, biometricians, epidemiologists, public health statisticians and social science researchers who have little or no knowledge of capture-recapture techniques.

How you will benefit

The course will develop your understanding of the huge potential of capture-recapture techniques.

Course content

The course is divided into two one-day sessions. Participants may attend one or both days.

Day 1: Capture-Recapture based upon Multiple Sources
Introduction, datasets and case studies
The classical two-sources Lincoln-Petersen and Lincoln-Petersen-Chapman estimates; underlying assumptions and limitations
Variance and confidence interval estimation
Extensions to more than two sources: log-linear modelling and its advantages

Day 2: Capture-Recapture based upon Continuous-Time Experiments
Introduction, datasets and case studies
Capture-recapture and zero-truncated count distributions
Chao lower bounds
Zelterman upper bounds
Variance and confidence interval estimation
Related techniques

Discounts

An Academic discount is available for this course.