13.07.2015 Views

Identifying Speculative Bubbles with an Infinite Hidden Markov Model

Identifying Speculative Bubbles with an Infinite Hidden Markov Model

Identifying Speculative Bubbles with an Infinite Hidden Markov Model

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

where n ji is the number of {τ | s τ = i, s τ−1 = j}.Sampling Θ | S, Y uses the results of regular linear models. The prior is:(ϕ i , σ −2i) ∼ NG(ϕ, H, χ 2 , ν 2 ).By conjugacy, the posterior is:<strong>with</strong>(ϕ i , σi−2 ) | S, Y ∼ NG(ϕ i , H i , χ i2 , ν i2 )ϕ i = H −1i (Hϕ + X ′ iY i )H i = H + X ′ iX iχ i = χ + Y ′i Y i + ϕ ′ Hϕ − ϕ ′ Hϕν i = ν + n iwhere Y i is the collection of y t in regime i. x t = (1, y t−1 , · · · , y t−q ). X i <strong>an</strong>d n i are the collectionof x t <strong>an</strong>d the number of observations in regime i, respectively.B.3 Sample (ϕ, H, χ) | S, Θ, νThe conditional posterior is:ϕ, H | {ϕ i , σ i } K i=1 ∼ NW(m 1 , τ 1 , A 1 , a 1 )30

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!