Improving the ‘samplev’ function, an efficient approach of sampling health states for individuals in microsimulation models

We have created a function that vectorizes the process of sampling from a multinomial distribution for microsimulation modeling. The most conventional way of running microsimulation models in R is to have two for loops: one for looping through the individuals and another for looping through the cycles in the time horizon. This is often quite computationally heavy when the number of individuals or cycles gets large. To improve the computational efficiency, we have created a function called ‘samplev’ that samples the health states of all individuals in the next cycle in one go from a multinomial distribution. This allows us to eliminate the for loop for the individuals. We further improved the efficiency by vectorizing the multinomial sampling process. The code for the function and a three-state microsimulation model that uses the function can be found here.