The video recording of the talk can be found at this link here.
To convert the Jupyter notebook to slides and serve in your browser:
- Install Jupyter if necessary
- Run
jupyter nbconvert slides.ipynb --to slides --post serveat the command line
Sophisticated tools are required to accurately estimate modern economic models, in the face of unprecedented macroeconomic conditions. The tempered particle filter is a novel method for filtering nonlinear state space models, surpassing conventional tools in accuracy and flexibility.
In this talk, I will highlight StateSpaceRoutines.jl, a repository containing state space filtering and smoothing methods such as the Kalman filter, Durbin Koopman smoother, and others. Specifically, I will discuss our latest addition to StateSpaceRoutines.jl, the “tempered particle filter” (TPF) which was developed by Ed Herbst (Federal Reserve Board of Governors) and Frank Schorfheide (University of Pennsylvania). TPF provides a robust way to simulate the distribution of the states and to calculate the likelihood in a general state space model. Furthermore, I will explain why TPF produces more accurate approximations than the standard bootstrap particle filter, and provide some results showing this. Because TPF is an inherently parallelizable algorithm, I will also delve into some details of the computational implementation and some lessons that we have learned along the way about Julia's parallel framework. I will highlight a few issues we are still currently facing with regards to optimizing the performance of TPF. StateSpaceRoutines.jl should prove useful to economists, statisticians, and those generally interested in Bayesian methods as a stand-alone suite of tools, which can be used to estimate a variety of both linear and non-linear state space models. Lastly, I will briefly highlight the NY Fed DSGE team's recent work on implementing various kinds of heterogeneous agent models, both in discrete and continuous time, and discuss the relevance of our implementations of TPF and other sequential Monte Carlo methods to estimating these models in the future.
This talk reflects the experience of the author and does not represent an endorsement by the Federal Reserve Bank of New York or the Federal Reserve System of any particular product or service. The views expressed in this talk are those of the author and do not necessarily reflect the position of the Federal Reserve Bank of New York or the Federal Reserve System. Any errors or omissions are the responsibility of the author.