The model structure is represented in the following graph. There are 4 possible states: 'Progression Free Survival in treatment'; 'Progression Free Survival off treatment'; 'Progressed disease' and 'Death'.
The traceplots can be used to visually inspect the Markov Chain Monte Carlo (MCMC) mixing process. We use two chains in the simulation; good mixing results in a 'fat hairy caterpillar' shape.
The Gelman-Rubin (GR) statistic, also known as Potential Scale Reduction (PSR) is a measure used to assess convergence of the MCMC process. Values of 1.1 and below indicate sufficient convergence.
The 'effective sample size' can be used to assess the level of autocorrelation. Ideally, the effective sample size should be close to the nominal sample size (i.e. the number of simulations in the MCMC process).
In MCMC simulations, by definition, autocorrelation occurs in subsequent lags and should approach values close to 0 with increasing lags.
Markov traces: progression-free survival (on and off treatment), progressed disease and death