SKATTNING AV INTENSITETER I EN OFULLSTÄNDIGT

              OBSERVERAD MARKOVPROCESS

            Nina Huzum och Emese Wieser

The purpose of this work is to estimate the incidence of
breast cancer relapse for those patients who have been
declared cured five years after a successful operation.
To achieve this goal we use a Markov model with continuous
time and four states. The transition intensities are
time-independent. On this basis we treat three different
cases characterized by various degrees of information
incompleteness. The required estimation is accomplished
with the ML method, in two of three cases using the EM
algorithm.