We want to develop methods of causal inference for processes that are followed over time. An important example is data on patients for clinical registries or cohorts (e.g. the Swiss HIV cohort with whom we have collaborated extensively).
The idea may be to estimate effects of treatment and diagnostic procedures in the absence of (or as a supplement to) randomized clinical trials. In particular we plan to cooperate with the Cancer Registry of Norway on this issue, using data from their recently developed quality registers of cancer patients.
A second issue is to understand mediators of treatment effects in randomized clinical trials. In medical survival studies much data go unanalyzed which could throw light on the ways which treatment effects take.
A third subject is to understand variation in risk between individuals, e.g. heterogeneity in cancer risk. This is partly based on family data