A regression-based capture-recapture approach with applications to life sciences (including medicine) and society

Research in the statistical sciences is strong at Southampton, with the Statistical Science Research Institute (S3RI) being a leader in variety of different areas of pure and applied statistical research. The statistical tools that are being developed in S3RI have huge potential applications in almost all fields across the life sciences. This project aims to apply the statistical method of “capture-recapture” to problems in the life sciences and medicine. The capture-recapture methodology is clearly only one area in statistics, but one that is receiving increasing attention outside the wildlife ecology area where it originally was developed. New areas of interest are surveillance of human populations for disease undercount or bioinformatics where the challenge is to estimate cellular “biodiversity”. One of the most theoretical challenges is the valid and asymptotically unbiased estimation of the size of the target population. The key idea of the current proposal is based upon a universal property inherent in many data sets from life and social sciences characterized by the fact that ratios of neighbouring frequencies follow a linear pattern. The key starting idea is the ratio plot in which the

ratios of neighbouring frequencies of capture-recapture counts are considered. This emerging pattern

of linearity can be illustrated in a number of applications of central interest including health sciences,

bioinformatics, security in society or surveillance studies. This (linear) pattern can be exploited to

develop various forms of regression-based estimators. A general theory is due to be developed to

show the validity of the estimating approach.