I have written several papers (Tanaka & Toda, 2013, 2015; Farmer & Toda, 2017; Toda, 2021) on discretizing probability distributions and stochastic processes.
I have compiled the discretization codes in a Matlab package that allows the user to:
- discretize a VAR(1) with Gaussian shocks,
- discretize an AR(p) process with Gaussian mixture shocks,
- discretize an AR(1) with log AR(1) stochastic volatility,
- discretize the Cox-Ingersoll-Ross model,
- calibrate a discrete distribution directly from data.
The package also contains a tutorial on how to write your own discretization codes. If you find any bugs or you have other processes that you would like to be discretized, please let me know by email.