# FFBS algorithm for estimating mean log-return parameter in stochastic volatility jump model

I am currently attempting to replicate this model: https://arxiv.org/pdf/1809.01501.pdf in r.

My (first) problem is regarding how to sample from conditional posterior for mu, $$(μ_{(j)}|Y_n, J_{(j−1)}, λ_{(j−1)}, γ_{(j−1)})$$. I have come across this http://hedibert.org/wp-content/uploads/2013/12/stochasticvolatilitymodels-R.txt r-code, describing similar process. But i cannot understand how to modify this so that i can include jumps in returns as described in the above paper.

The main problem is that i do not understand how to practically execute a FFBS algorithm properly with regards to what the variables in the algorithm do.

Could someone atleast try to point me in the right direction? thanks

• Welcome to CV! In addition to linking those sources, could you brief describe them? This would help make your question specific and self-contained. – eric_kernfeld Jun 25 '19 at 14:44
• Thank you! i wish i could describe exactly what it is i do not understand and i do not know the mathjax (or whatever its called) syntax. But i will certainly revise it so that it is so specific as i can. – Victor Tisell Jun 25 '19 at 15:05
• I formatted the math -- does that look right to you? – eric_kernfeld Jun 25 '19 at 17:15
• I can sympathize with the sentiment that descriptions are often opaque in papers like this, but the reason is that it is so well known. The term you want to google for is “simulation smoothing”, which will provide you with more details. A central paper is also Kim, Shephard and Chib (1998). – hejseb Jun 25 '19 at 18:14
• @eric_kernfeld, looks excellent. – Victor Tisell Jun 25 '19 at 18:38