Nonparametric estimation of transition probabilities in non-Markov multi-state models
Research Objectives:
In many instances, survival data are observed under complicated observational biasing schemes. This is the case, for example, when considering (rather than incident) prevalent cases, which are sampled in a cross-sectional way. Another example is found in medical registries, when they only provide information on patients who have been diagnosed between two specific dates. In general, different combinations of truncation and censoring patterns will affect the recruited lifetimes, and proper corrections are needed for estimation and inference purposes. The main goal of this project is the adjustment of statistical methods to correct the referred observational biases, so one can provide reliable estimation of population targets such as the mean lifetime, the survival function, or the regression function. Suitable testing methods for groups or treatments comparison will be a focus of attention too.
Also have a look at the research group of Jacobo de Uña Álvarez.
Description of work:
Besides the technical results, it is expected that some friendly (free) software will be developed, so that practitioners can apply the proposed methods in a simple manner. The methodology is, e.g., relevant for summarizing survival rates from cancer registries, and for performing group comparisons, which may help to identify risk factors. Another application will be fitting algorithms for models with explanatory (e.g. treatment) variables in survival prognosis.
Host Institution: University of Vigo