# C the prior distribution for μ and μ1 is normal μ

• No School
• AA 1
• 60

This preview shows page 45 - 48 out of 60 pages.

c. The prior distribution for μ 0 and μ 1 is normal: μ 0 μ 1 ∼ N ( a, V ) . Define Y * = { y * t } , such that y * t = f i t - s i t × x i t , and X * = [ 1 - S i , S i ]. Then, the posterior distribution is also normal, N a, ¯ V ), such that ¯ V = V - 1 + ( X * ) 0 X * - 1 ; ¯ a = ¯ V V - 1 a + ( X * ) 0 Y * . The regime-specific means μ i +1 0 and μ i +1 1 are simulated from this posterior. d. In our specifications, we have worked with only one quarterly variable, GDP growth. We fix the common-factor loading γ 1 for the quarterly GDP growth to be equal to one for identification purposes—this assumption amounts to scaling the common factor. Then, the factor loadings for the monthly variables are interpreted as relative to the unit loading for the GDP growth. In the prior, a factor loading γ j of a monthly indicator j = 1 , ..., M is normally distributed: γ j ∼ N ( a, V ). Then, define ˜ y j,t and ˜ f j,t as follows: ˜ y j,t = y j,t - ψ i j, 1 y j,t - 1 - ... - ψ i j,P y j,t - P ; ˜ f i j,t = f i t - ψ i j, 1 f i t - 1 - ... - ψ i j,P f i t - P . We can use these definitions together with equations ( A.3 ) and ( A.4 ) to derive the following expression: ˜ y j,t = γ j ˜ f i j,t + e j,t , e j,t ∼ N (0 , σ 2 j ) . Using this expression, we can find the posterior for the factor loading to be normal ECB Working Paper Series No 2381 / March 2020 44
as well: γ j ∼ N a j , ¯ V j ), such that ¯ V j = V - 1 + ( ˜ X j ) 0 ˜ X j - 1 ; ¯ a j = ¯ V V - 1 a + ( ˜ X j ) 0 ˜ Y j , where ˜ X j and ˜ Y j are vectors with elements { ˜ f j,t } and { ˜ y j,t } defined above. Factor loadings { γ i +1 j } are simulated from these posteriors. e. We assume that the individual component of the only quarterly variable in our model is a white noise, which makes it simpler to compute the posterior, due to the monthly missing observations in a variable at the quarterly frequency: for the individual com- ponent of GDP growth, equation ( A.4 ) reduces to u 1 ,t = e 1 ,t , e 1 ,t ∼ N (0 , σ 2 1 ) . Then, we specify the inverse-gamma prior σ 2 1 ∼ IG ( a, b ), which conjugates with the inverse-gamma posterior IG a, ¯ b ), such that ¯ a = a + T 2 , 1 b + ( y i 1 ,t ) 2 / 2 . The variance ( σ i +1 1 ) 2 is simulated from this posterior. f. Define Y j = ( y j, 1 , ..., y j,T ) to be the vector collecting the observations of a monthly variable j . Let X j be the T × P matrix recording the P lags of the variable y j,t . For the AR coefficients of each monthly variable’s individual component, ψ j = ( ψ j, 1 , ..., ψ j,P ) 0 , we assume the same normal prior: ψ j ∼ N ( a, V ). Then, the posterior is also normal, N a j , ¯ V j ), but different for each j , such that ¯ V j = V - 1 a + X 0 j X j ( σ i j ) 2 ! - 1 ; ¯ a j = ¯ V j V - 1 a + X 0 j Y j ( σ i j ) 2 ! We simulate ψ i +1 j from this posterior. Finally, to simulate σ i +1 j , we assume inverse- ECB Working Paper Series No 2381 / March 2020 45
gamma prior: for each j , σ 2 j ∼ IG ( a, b ). Let Ψ i +1 j be the following P × P matrix: Ψ i +1 j = ψ i +1 j, 1 · · · ψ i +1 j,P - 1 ψ i +1 j,P 1 · · · 0 0 . . . . . . . . . . . . 0 · · · 1 0 Then, we can express the posterior of σ 2 j as inverse-gamma distribution as well, IG a, ¯ b ), such that ¯ a = a + T 2 ; ¯ b = b + ( Y j - X j Ψ i +1 j ) 0 ( Y j - X j Ψ i +1 j ) 2 !

#### You've reached the end of your free preview.

Want to read all 60 pages?

• Fall '19
• Economics, Recession, Late-2000s recession, GWI

### What students are saying

• As a current student on this bumpy collegiate pathway, I stumbled upon Course Hero, where I can find study resources for nearly all my courses, get online help from tutors 24/7, and even share my old projects, papers, and lecture notes with other students.

Kiran Temple University Fox School of Business ‘17, Course Hero Intern

• I cannot even describe how much Course Hero helped me this summer. It’s truly become something I can always rely on and help me. In the end, I was not only able to survive summer classes, but I was able to thrive thanks to Course Hero.

Dana University of Pennsylvania ‘17, Course Hero Intern

• The ability to access any university’s resources through Course Hero proved invaluable in my case. I was behind on Tulane coursework and actually used UCLA’s materials to help me move forward and get everything together on time.

Jill Tulane University ‘16, Course Hero Intern