G. Gómez Melis, J. Najera-Zuloaga, M. Besalu Mayol
Multistate models can be used to describe the movement of patients among different states. While first order Markov condition where the future evolution only depends on the current state might be too restrictive, second order Markov could provide a more realistic description of the reality. Second order Markov models assume that the progression of the patient not only depends on the current state but also the preceding. Inference in this case is not straightforward because it entails a MxMxM matrix (M is the number of states). An extended transition probability matrix as M different matrices of order MxM has been proposed, as well as an extension of the Chapman-Kolmogorov equations.
In this talk we present two different nonparametric estimators for the transition probabilities and explore their properties both asymptotically and by simulation. The methods will be illustrated for COVID-19 patients at a high risk of developing severe outcomes.
Keywords: Multi-State Models, Non-Markov
Scheduled
Biostatistics I
November 8, 2023 12:00 PM
CC2: Conference Room