Prediction is very difficult, especially if it’s about the future. — Neils Bohr (玻尔).

 

I am working on building some R and Python packages for both R and Python. The packages are mostly focused on financial time series models such as interest rate models, stochastic volatility models, stochastic conditional duration models, latent factor models including non-linear and non-normal state space models, probabilistic principal component models, and component factor models etc. The final goal is to build a set of packages to tackle those models so people can use them to solve the problems without re-implement the estimation algorithms.

As these models are time-dependent and sometimes are overparameterized, usual MLE may not help to fit those models. We develop Markov Chain Monte Carlo (MCMC) methods within Bayesian framework. As simulation-based inferences are time-consuming, we first do experimentation in R and Python, and then build packages by rewriting the algorithms in C/C++ to speed up the estimation process.

Besides on package building, I will also write tutorials to introduce basic concepts of simulation-based inference such as acceptance-rejection methods, random number generation for non-usual distributions, slice sampler, Metropolis_Hastings methods etc. Those methods are just to get you familiar on why and how we need simulation based inference.

This is a long journey, please visit us frequently to check the updates.

If you have any comments or suggestions, please kindly let me know by sending me an email to: