Functional Data Analysis for the investigation of longitudinal response patterns in health care

A total of 137 subjects are followed over a year with two weekly intervals. At each measurement the presence or absence of a specific pathogen is noted, and if the pathogen is present, the specific serotype is determined. The aim of the study is to investigate patterns of colonisation and relationships with disease. Functional data analysis is used to model the occurrence of serotypes. Furthermore the subjects are vaccinated at three occasions during the study. Only a few of the serotypes are covered by the vaccine. A complicating factor is that the vaccinations did not all take place at the same point in time for all the subjects. The use of functional data analysis and specifically curve registration to ‘align’ the vaccination points is illustrated in the comparison of colonisation patterns in response to the vaccinations.