News | July 31, 2012
MPIDR researcher Ekaterina Ogurtsova was awarded a prize for the Best Student Presentation at the 27th International Workshop on Statistical Modelling held in Prague July 16-20. Ekaterina, who is a PhD student in the Laboratory for Statistical Demography, reported results from her doctoral research in a presentation titled "Estimation of Multi-State Model Parameters from Panel Data: A Comparison of Different Methods".
Multi-state models are an important tool if one wants to study how people pass through different stages in their lives. This can be questions of when we make the transition from being healthy to a state where we suffer from certain health problems until we finally move on to death. Or, in more positive terms, how people choose the way they live after having left the parental home: Do they live alone and if so, for how long? When do they start to live with a partner i.e. cohabitate? When do they get married, and perhaps later divorced? Multi-state models allow researchers to study as well as project prevalences of certain diseases, of healthy life expectancy or the average length of marriages.
A problem in the application of such models is that empirical data often are not as complete as statisticians would like them to be. Ideally, we would like to observe how and when, and at which age, people make all the relevant transitions. In practice, however, we often only know in which state people are at the time of interview, which may be only every year or every other year - a scheme called panel data. So we may miss some recovery from bad to good health, or we may miss a phase of cohabitation just because cohabitation took place in-between two interviews. Statisticians have several methods to address this problem. In her thesis, Ekaterina Ogurtsova compares some of these methods and assesses which of them is to be preferred in practical applications.