bmgarch estimates Bayesian multivariate generalized autoregressive conditional heteroskedasticity (MGARCH) models. Currently, bmgarch supports ARMA(1,1) and intercept-only (Constant) mean structures, and a variety of MGARCH(P,Q) parameterizations. In increasing order of complexity:
bmgarch is not yet available on CRAN.
The development version can be installed from GitHub with:
devtools::install_github("ph-rast/bmgarch")We present two examples, one with behavioral data and one with stocks from three major Japanese automakers.
In this example, we use the pdBEKK(1,1) model for the variances, and an intercept-only model for the means.
library(bmgarch)
data(panas)
head(panas)
#> Pos Neg
#> 1 -2.193 -2.419
#> 2 1.567 -0.360
#> 3 -0.124 -1.202
#> 4 0.020 -1.311
#> 5 -0.150 2.004
#> 6 3.877 1.008
# Fit pdBEKK(1, 1) with ARMA(1,1) on the mean structure.
fit <- bmgarch(panas,
parameterization = "pdBEKK",
iterations = 1000,
P = 1, Q = 1,
distribution = "Student_t",
meanstructure = "arma")
#>
#> CHECKING DATA AND PREPROCESSING FOR MODEL 'pdBEKKMGARCH' NOW.
#>
#> COMPILING MODEL 'pdBEKKMGARCH' NOW.
#>
#> STARTING SAMPLER FOR MODEL 'pdBEKKMGARCH' NOW.summary(fit)
#> Model: pdBEKK-MGARCH
#> Basic Specification: H_t = D_t R D_t
#> H_t = C + A'[y_(t-1)*y'_(t-1)]A + B'H_(t-1)B
#>
#> Distribution: Student_t
#> ---
#> Iterations: 1000
#> Chains: 4
#> Date: Tue Sep 1 22:06:52 2020
#> Elapsed time (min): 14.78
#>
#> ---
#> Constant correlation, R (diag[C]*R*diag[C]):
#>
#> mean sd mdn 2.5% 97.5% n_eff Rhat
#> R_Ng-Ps 0.01 0.49 0 -0.88 0.9 1225.16 1
#>
#>
#> Constant variances (diag[C]):
#>
#> mean sd mdn 2.5% 97.5% n_eff Rhat
#> var_Ps 0.77 0.98 0.34 0.01 3.53 277.21 1
#> var_Ng 1.20 0.45 1.27 0.16 1.95 361.59 1
#>
#>
#> MGARCH(1,1) estimates for A:
#>
#> mean sd mdn 2.5% 97.5% n_eff Rhat
#> A_Ps-Ps 0.34 0.10 0.34 0.13 0.55 1236.60 1
#> A_Ng-Ps 0.06 0.08 0.06 -0.09 0.21 1539.86 1
#> A_Ps-Ng 0.05 0.14 0.06 -0.23 0.34 1001.81 1
#> A_Ng-Ng 0.39 0.12 0.39 0.13 0.61 984.89 1
#>
#>
#> MGARCH(1,1) estimates for B:
#>
#> mean sd mdn 2.5% 97.5% n_eff Rhat
#> B_Ps-Ps 0.71 0.25 0.83 0.05 0.95 232.08 1.01
#> B_Ng-Ps -0.09 0.18 -0.09 -0.52 0.33 252.72 1.01
#> B_Ps-Ng 0.27 0.40 0.27 -0.67 1.16 388.01 1.01
#> B_Ng-Ng 0.35 0.21 0.34 0.02 0.78 785.91 1.00
#>
#>
#> ARMA(1,1) estimates on the location:
#>
#> mean sd mdn 2.5% 97.5% n_eff Rhat
#> (Intercept)_Pos -0.03 0.17 -0.02 -0.42 0.29 683.62 1.00
#> (Intercept)_Neg 0.07 0.12 0.06 -0.16 0.34 949.69 1.01
#> Phi_Pos-Pos -0.06 0.39 -0.03 -0.87 0.67 577.24 1.00
#> Phi_Pos-Neg -0.10 0.46 -0.12 -0.89 0.81 534.97 1.00
#> Phi_Neg-Pos -0.16 0.36 -0.16 -0.85 0.62 626.77 1.01
#> Phi_Neg-Neg 0.05 0.43 0.06 -0.71 0.82 858.42 1.00
#> Theta_Pos-Pos -0.02 0.41 -0.04 -0.79 0.82 574.07 1.00
#> Theta_Pos-Neg 0.01 0.47 0.05 -0.92 0.79 506.37 1.00
#> Theta_Neg-Pos 0.18 0.36 0.18 -0.61 0.86 595.10 1.02
#> Theta_Neg-Neg -0.06 0.44 -0.06 -0.86 0.74 810.21 1.00
#>
#>
#> Df constant student_t (nu):
#>
#> mean sd mdn 2.5% 97.5% n_eff Rhat
#> 52.18 28.07 47.06 13.69 119.03 2529.13 1.00
#>
#>
#> Log density posterior estimate:
#>
#> mean sd mdn 2.5% 97.5% n_eff Rhat
#> -803.41 4.97 -803.11 -814.06 -794.68 383.58 1.00fit.fc <- forecast(fit, ahead = 5)
fit.fc
#> ---
#> [Mean] Forecast for 5 ahead:
#>
#> Pos :
#>
#> period mean sd mdn 2.5% 97.5% n_eff Rhat
#> 201 -0.25 2.91 -0.29 -6.00 5.46 1860.08 1
#> 202 -0.19 2.77 -0.20 -5.60 5.41 2037.99 1
#> 203 -0.13 2.75 -0.08 -5.68 4.97 1952.89 1
#> 204 -0.14 2.59 -0.16 -5.24 5.03 1952.77 1
#> 205 -0.09 2.60 -0.09 -5.20 5.28 1870.26 1
#> Neg :
#>
#> period mean sd mdn 2.5% 97.5% n_eff Rhat
#> 201 0.35 1.45 0.36 -2.53 3.16 1698.74 1
#> 202 0.25 1.58 0.22 -2.72 3.54 1594.92 1
#> 203 0.15 1.59 0.14 -3.14 3.26 1923.31 1
#> 204 0.20 1.61 0.25 -3.01 3.33 2060.69 1
#> 205 0.14 1.57 0.10 -2.98 3.24 2014.93 1
#> ---
#> [Variance] Forecast for 5 ahead:
#>
#> Pos :
#>
#> period mean sd mdn 2.5% 97.5% n_eff Rhat
#> 201 7.77 2.61 7.48 3.89 13.49 570.52 1
#> 202 7.16 3.28 6.52 3.36 15.00 898.08 1
#> 203 6.73 3.73 5.82 3.11 15.35 1172.30 1
#> 204 6.51 4.63 5.33 3.04 16.12 1578.67 1
#> 205 6.29 4.65 5.08 2.93 16.37 1613.58 1
#> Neg :
#>
#> period mean sd mdn 2.5% 97.5% n_eff Rhat
#> 201 1.97 0.43 1.90 1.36 3.02 1031.77 1
#> 202 2.22 0.76 2.03 1.40 4.11 2006.64 1
#> 203 2.31 0.99 2.07 1.41 4.62 2062.82 1
#> 204 2.34 1.00 2.08 1.41 4.97 2096.10 1
#> 205 2.31 1.08 2.05 1.40 4.72 2017.50 1
#> [Correlation] Forecast for 5 ahead:
#>
#> Neg_Pos :
#>
#> period mean sd mdn 2.5% 97.5% n_eff Rhat
#> 201 -0.07 0.17 -0.06 -0.40 0.23 1012.83 1
#> 202 -0.07 0.21 -0.07 -0.46 0.37 1326.89 1
#> 203 -0.06 0.21 -0.06 -0.50 0.39 1619.66 1
#> 204 -0.04 0.21 -0.04 -0.46 0.41 1721.83 1
#> 205 -0.03 0.20 -0.04 -0.45 0.41 1537.11 1
plot(fit.fc, askNewPage = FALSE, type = "var")
plot(fit.fc, askNewPage = FALSE, type = "cor")
Here we use the first 100 days of Stata’s stocks data on daily lagged returns of three Japanese automakers, Toyota, Nissan, and Honda.
library(bmgarch)
data(stocks)
head(stocks)
#> date t toyota nissan honda
#> 1 2003-01-02 1 0.015167475 0.029470444 0.031610250
#> 2 2003-01-03 2 0.004820108 0.008173466 0.002679110
#> 3 2003-01-06 3 0.019958735 0.013064146 -0.001606464
#> 4 2003-01-07 4 -0.013322592 -0.007444382 -0.011317968
#> 5 2003-01-08 5 -0.027001143 -0.018856525 -0.016944885
#> 6 2003-01-09 6 0.011634588 0.016986847 0.013687611Ease computation by first standardizing the time series
stocks.z <- scale(stocks[,c("toyota", "nissan", "honda")])
head(stocks.z )
#> toyota nissan honda
#> 1 0.8151655 1.3417896 1.52836901
#> 2 0.2517820 0.3687089 0.11213515
#> 3 1.0760354 0.5921691 -0.09765177
#> 4 -0.7360344 -0.3448866 -0.57304819
#> 5 -1.4807910 -0.8663191 -0.84849638
#> 6 0.6228102 0.7714013 0.65102202
# Fit CCC(1, 1) with constant on the mean structure.
fit1 <- bmgarch(stocks.z[1:100, c("toyota", "nissan", "honda")],
parameterization = "CCC",
iterations = 1000,
P = 1, Q = 1,
distribution = "Student_t",
meanstructure = "constant")
#>
#> CHECKING DATA AND PREPROCESSING FOR MODEL 'CCCMGARCH' NOW.
#>
#> COMPILING MODEL 'CCCMGARCH' NOW.
#>
#> STARTING SAMPLER FOR MODEL 'CCCMGARCH' NOW.summary( fit1 )
#> Model: CCC-MGARCH
#> Basic Specification: H_t = D_t R D_t
#> diag(D_t) = sqrt(h_[ii,t]) = c_h + a_h*y^2_[t-1] + b_h*h_[ii, t-1
#>
#> Distribution: Student_t
#> ---
#> Iterations: 1000
#> Chains: 4
#> Date: Tue Sep 1 22:08:07 2020
#> Elapsed time (min): 0.95
#>
#> GARCH(1,1) estimates for conditional variance:
#>
#> mean sd mdn 2.5% 97.5% n_eff Rhat
#> a_h_1,ty 0.11 0.10 0.08 0.00 0.37 2185.86 1
#> a_h_1,ns 0.08 0.07 0.06 0.00 0.26 2744.57 1
#> a_h_1,hn 0.11 0.09 0.09 0.00 0.33 2377.30 1
#> b_h_1,ty 0.45 0.18 0.46 0.10 0.75 1616.31 1
#> b_h_1,ns 0.37 0.20 0.35 0.06 0.77 1373.82 1
#> b_h_1,hn 0.39 0.18 0.38 0.08 0.76 1316.46 1
#> c_h_var_ty 0.29 0.12 0.27 0.10 0.56 1455.06 1
#> c_h_var_ns 0.35 0.13 0.36 0.10 0.60 1527.56 1
#> c_h_var_hn 0.44 0.16 0.43 0.16 0.78 1264.84 1
#>
#>
#> Constant correlation (R) coefficients:
#>
#> mean sd mdn 2.5% 97.5% n_eff Rhat
#> R_ns-ty 0.65 0.07 0.65 0.50 0.76 2426.85 1
#> R_hn-ty 0.73 0.05 0.74 0.62 0.82 2109.22 1
#> R_hn-ns 0.64 0.07 0.64 0.50 0.75 2336.69 1
#>
#>
#> Intercept estimates on the location:
#>
#> mean sd mdn 2.5% 97.5% n_eff Rhat
#> (Intercept)_toyota -0.09 0.08 -0.09 -0.24 0.07 1729.36 1
#> (Intercept)_nissan -0.01 0.08 -0.01 -0.16 0.15 1693.25 1
#> (Intercept)_honda -0.02 0.09 -0.02 -0.20 0.16 1574.42 1
#>
#>
#> Df constant student_t (nu):
#>
#> mean sd mdn 2.5% 97.5% n_eff Rhat
#> 31.68 23.48 24.19 7.09 91.31 2270.38 1.00
#>
#>
#> Log density posterior estimate:
#>
#> mean sd mdn 2.5% 97.5% n_eff Rhat
#> -178.37 5.22 -178.00 -189.56 -169.32 705.81 1.00Forecast volatility 10 days ahead
fc <- forecast(fit1, ahead = 10 )
fc
#> ---
#> [Variance] Forecast for 10 ahead:
#>
#> toyota :
#>
#> period mean sd mdn 2.5% 97.5% n_eff Rhat
#> 101 0.54 0.11 0.53 0.34 0.77 1652.55 1
#> 102 0.58 0.17 0.56 0.35 0.96 1983.69 1
#> 103 0.61 0.22 0.58 0.36 1.09 1950.83 1
#> 104 0.63 0.24 0.59 0.37 1.15 1862.93 1
#> 105 0.65 0.29 0.60 0.37 1.27 2011.49 1
#> 106 0.66 0.36 0.60 0.37 1.38 1816.63 1
#> 107 0.67 0.33 0.61 0.38 1.38 1910.15 1
#> 108 0.67 0.37 0.61 0.38 1.36 1816.65 1
#> 109 0.68 0.32 0.61 0.38 1.35 2084.51 1
#> 110 0.67 0.33 0.61 0.39 1.31 1920.16 1
#> nissan :
#>
#> period mean sd mdn 2.5% 97.5% n_eff Rhat
#> 101 0.61 0.10 0.60 0.43 0.84 2066.22 1
#> 102 0.64 0.15 0.62 0.43 0.97 1752.66 1
#> 103 0.66 0.18 0.62 0.43 1.05 1942.56 1
#> 104 0.67 0.21 0.63 0.43 1.11 1810.40 1
#> 105 0.67 0.23 0.63 0.43 1.20 1880.72 1
#> 106 0.67 0.24 0.63 0.43 1.22 1737.02 1
#> 107 0.68 0.23 0.63 0.43 1.23 1954.45 1
#> 108 0.68 0.23 0.63 0.43 1.23 1931.12 1
#> 109 0.68 0.27 0.63 0.44 1.16 1635.07 1
#> 110 0.68 0.23 0.64 0.43 1.19 2138.37 1
#> honda :
#>
#> period mean sd mdn 2.5% 97.5% n_eff Rhat
#> 101 0.76 0.14 0.75 0.53 1.07 1787.36 1
#> 102 0.82 0.21 0.79 0.53 1.29 1905.72 1
#> 103 0.86 0.26 0.81 0.54 1.44 1636.53 1
#> 104 0.89 0.34 0.82 0.55 1.58 1658.82 1
#> 105 0.92 0.70 0.83 0.55 1.67 1346.66 1
#> 106 0.93 1.16 0.84 0.56 1.78 1752.68 1
#> 107 0.96 2.27 0.84 0.56 1.79 1920.54 1
#> 108 0.95 1.43 0.84 0.56 1.84 1973.01 1
#> 109 0.94 0.59 0.84 0.55 1.85 2077.14 1
#> 110 0.95 0.88 0.84 0.56 1.98 2018.69 1
plot(fc,askNewPage = FALSE, type = 'var' )
Here we illustrate how to obtain model weights across three models. These weights will be used to compute weighted forecasts, thus, taking into account that we do not have a single best model.
Add two additional models, one with CCC(2,2) and a DCC(1,1)
# Fit CCC(1, 1) with constant on the mean structure.
fit2 <- bmgarch(stocks.z[1:100, c("toyota", "nissan", "honda")],
parameterization = "CCC",
iterations = 1000,
P = 2, Q = 2,
distribution = "Student_t",
meanstructure = "constant")
#>
#> CHECKING DATA AND PREPROCESSING FOR MODEL 'CCCMGARCH' NOW.
#>
#> COMPILING MODEL 'CCCMGARCH' NOW.
#>
#> STARTING SAMPLER FOR MODEL 'CCCMGARCH' NOW.
fit3 <- bmgarch(stocks.z[1:100, c("toyota", "nissan", "honda")],
parameterization = "DCC",
iterations = 1000,
P = 1, Q = 1,
distribution = "Student_t",
meanstructure = "arma")
#>
#> CHECKING DATA AND PREPROCESSING FOR MODEL 'DCCMGARCH' NOW.
#>
#> COMPILING MODEL 'DCCMGARCH' NOW.
#>
#> STARTING SAMPLER FOR MODEL 'DCCMGARCH' NOW.The DCC(1,1) model also incorporates an ARMA(1,1) meanstructure. The output will have the according information:
summary( fit3 )
#> Model: DCC-MGARCH
#> Basic Specification: H_t = D_t R D_t
#> diag(D_t) = sqrt(h_ii,t) = c_h + a_h*y^2_[t-1] + b_h*h_[ii,t-1]
#> R_t = Q^[-1]_t Q_t Q^[-1]_t = ( 1 - a_q - b_q)S + a_q(u_[t-1]u'_[t-1]) + b_q(Q_[t-1])
#>
#> Distribution: Student_t
#> ---
#> Iterations: 1000
#> Chains: 4
#> Date: Tue Sep 1 22:24:48 2020
#> Elapsed time (min): 15.41
#>
#> GARCH(1,1) estimates for conditional variance on D:
#>
#> mean sd mdn 2.5% 97.5% n_eff Rhat
#> a_h_1,ty 0.15 0.13 0.14 0.02 0.44 4.90 1.31
#> a_h_1,ns 0.07 0.07 0.07 0.01 0.27 8.17 1.17
#> a_h_1,hn 0.16 0.11 0.11 0.01 0.33 3.41 1.54
#> b_h_1,ty 0.42 0.12 0.42 0.15 0.70 49.76 1.05
#> b_h_1,ns 0.54 0.19 0.62 0.09 0.80 4.75 1.32
#> b_h_1,hn 0.48 0.14 0.46 0.15 0.79 155.25 1.04
#> c_h_var_ty 0.33 0.12 0.29 0.11 0.50 3.50 1.52
#> c_h_var_ns 0.26 0.11 0.25 0.10 0.54 6.86 1.21
#> c_h_var_hn 0.32 0.13 0.29 0.14 0.69 10.34 1.13
#>
#>
#> GARCH(1,1) estimates for conditional variance on Q:
#>
#> mean sd mdn 2.5% 97.5% n_eff Rhat
#> a_q 0.29 0.15 0.22 0.05 0.51 2.49 2.20
#> b_q 0.29 0.11 0.30 0.04 0.48 7.21 1.18
#>
#>
#> Unconditional correlation 'S' in Q:
#>
#> mean sd mdn 2.5% 97.5% n_eff Rhat
#> S_ns-ty 0.64 0.08 0.62 0.46 0.74 3.52 1.51
#> S_hn-ty 0.72 0.07 0.74 0.61 0.82 3.56 1.50
#> S_hn-ns 0.59 0.08 0.61 0.49 0.74 4.04 1.40
#>
#>
#> ARMA(1,1) estimates on the location:
#>
#> mean sd mdn 2.5% 97.5% n_eff Rhat
#> (Intercept)_toyota -0.06 0.07 -0.07 -0.22 0.07 12.86 1.12
#> (Intercept)_nissan 0.11 0.12 0.11 -0.13 0.28 2.82 1.84
#> (Intercept)_honda -0.02 0.09 -0.01 -0.19 0.21 116.92 1.04
#> Phi_toyota-toyota -0.01 0.36 0.04 -0.63 0.63 3.92 1.44
#> Phi_toyota-nissan -0.17 0.40 -0.02 -0.69 0.62 3.66 1.49
#> Phi_toyota-honda 0.23 0.32 0.20 -0.56 0.70 6.65 1.23
#> Phi_nissan-toyota 0.47 0.33 0.61 -0.44 0.87 9.28 1.17
#> Phi_nissan-nissan -0.32 0.29 -0.46 -0.76 0.45 11.13 1.15
#> Phi_nissan-honda -0.02 0.41 0.10 -0.63 0.72 3.93 1.45
#> Phi_honda-toyota -0.17 0.38 -0.25 -0.90 0.45 4.93 1.32
#> Phi_honda-nissan -0.06 0.42 -0.01 -0.57 0.86 3.77 1.46
#> Phi_honda-honda -0.29 0.31 -0.37 -0.71 0.47 5.58 1.28
#> Theta_toyota-toyota -0.16 0.37 -0.14 -0.76 0.52 4.45 1.37
#> Theta_toyota-nissan 0.34 0.41 0.19 -0.50 0.87 3.51 1.52
#> Theta_toyota-honda -0.14 0.33 -0.08 -0.68 0.56 5.21 1.30
#> Theta_nissan-toyota -0.49 0.34 -0.64 -0.88 0.47 7.64 1.20
#> Theta_nissan-nissan 0.34 0.28 0.46 -0.37 0.74 8.30 1.18
#> Theta_nissan-honda -0.02 0.43 -0.15 -0.80 0.56 3.34 1.59
#> Theta_honda-toyota -0.20 0.40 -0.06 -0.77 0.64 4.32 1.38
#> Theta_honda-nissan 0.13 0.41 0.06 -0.79 0.69 4.71 1.34
#> Theta_honda-honda 0.57 0.44 0.89 -0.45 0.91 3.28 1.61
#>
#>
#> Df constant student_t (nu):
#>
#> mean sd mdn 2.5% 97.5% n_eff Rhat
#> 36.72 22.41 35.72 11.66 100.92 8.59 1.15
#>
#>
#> Log density posterior estimate:
#>
#> mean sd mdn 2.5% 97.5% n_eff Rhat
#> -171.51 8.38 -172.71 -187.27 -158.23 2.50 2.24
fc <- forecast(fit3, ahead = 10)
plot( fc,askNewPage = FALSE, type = 'mean' ) 
Obtain model weights with either the stacking or the pseudo BMA method. These methods are inherited from the loo package.
First, gather models to a bmgarch_list.
## use bmgarch_list function to collect bmgarch objects
modfits <- bmgarch_list(fit1, fit2, fit3)Compute model weights with the stacking method (default) and the the approximate (default) leave-future-out cross validation (LFO CV). L defines the minimal length of the time series before we start engaging in cross-validation. Eg., for a time series with length 100, L = 50 reserves values 51–100 as the cross-validation sample. Note that the standard is to use the approximate backward method to CV as it results in fewest refits. Exact CV is also available with exact but not encouraged as it results in refitting all CV models.
mw <- model_weights(modfits, L = 50, method = 'stacking')
#> Using threshold 0.6 , model was refit 10 times, at observations 93 85 83 75 74 72 64 60 58 52
## Return model weights:
mw
#> Method: stacking
#> ------
#> weight
#> model1 0.555
#> model2 0.445
#> model3 0.000Use model weights to obtain weighted forecasts. Here we will forecast 5 days ahead.
w_fc <- forecast(modfits, ahead = 5, weights = mw )
w_fc
#> ---
#> LFO-weighted forecasts across 3 models.
#> ---
#> [Mean] Forecast for 5 ahead:
#>
#> toyota :
#>
#> period mean sd mdn 2.5% 97.5% n_eff Rhat
#> 101 -0.09 0.57 -0.09 -1.23 1.01 NA NA
#> 102 -0.08 0.59 -0.09 -1.23 1.13 NA NA
#> 103 -0.10 0.62 -0.09 -1.35 1.13 NA NA
#> 104 -0.10 0.61 -0.09 -1.37 1.11 NA NA
#> 105 -0.08 0.62 -0.07 -1.31 1.13 NA NA
#> nissan :
#>
#> period mean sd mdn 2.5% 97.5% n_eff Rhat
#> 101 0.00 0.62 -0.01 -1.22 1.22 NA NA
#> 102 0.00 0.65 0.02 -1.31 1.24 NA NA
#> 103 -0.01 0.64 0.00 -1.26 1.26 NA NA
#> 104 0.00 0.64 -0.01 -1.33 1.29 NA NA
#> 105 0.01 0.63 0.02 -1.21 1.27 NA NA
#> honda :
#>
#> period mean sd mdn 2.5% 97.5% n_eff Rhat
#> 101 -0.04 0.69 -0.04 -1.39 1.31 NA NA
#> 102 -0.02 0.70 0.00 -1.39 1.35 NA NA
#> 103 -0.05 0.74 -0.03 -1.54 1.40 NA NA
#> 104 -0.02 0.75 -0.02 -1.53 1.50 NA NA
#> 105 0.00 0.73 0.00 -1.44 1.43 NA NA
#> ---
#> [Variance] Forecast for 5 ahead:
#>
#> toyota :
#>
#> period mean sd mdn 2.5% 97.5% n_eff Rhat
#> 101 0.53 0.08 0.53 0.39 0.70 NA NA
#> 102 0.57 0.11 0.55 0.40 0.82 NA NA
#> 103 0.60 0.16 0.58 0.41 0.97 NA NA
#> 104 0.62 0.18 0.59 0.41 1.04 NA NA
#> 105 0.64 0.20 0.60 0.42 1.13 NA NA
#> nissan :
#>
#> period mean sd mdn 2.5% 97.5% n_eff Rhat
#> 101 0.62 0.08 0.61 0.48 0.80 NA NA
#> 102 0.64 0.12 0.63 0.48 0.89 NA NA
#> 103 0.66 0.14 0.64 0.48 1.00 NA NA
#> 104 0.67 0.16 0.64 0.48 1.02 NA NA
#> 105 0.67 0.16 0.65 0.48 1.04 NA NA
#> honda :
#>
#> period mean sd mdn 2.5% 97.5% n_eff Rhat
#> 101 0.77 0.10 0.76 0.60 1.00 NA NA
#> 102 0.81 0.16 0.79 0.58 1.14 NA NA
#> 103 0.87 0.22 0.83 0.61 1.38 NA NA
#> 104 0.89 0.25 0.84 0.61 1.50 NA NA
#> 105 0.91 0.28 0.86 0.62 1.57 NA NA
#> [Correlation] Forecast for 5 ahead:
#>
#> nissan_toyota :
#>
#> period mean sd mdn 2.5% 97.5% n_eff Rhat
#> 101 0.65 0.05 0.65 0.55 0.73 NA NA
#> 102 0.65 0.05 0.65 0.55 0.73 NA NA
#> 103 0.65 0.05 0.65 0.55 0.73 NA NA
#> 104 0.65 0.05 0.65 0.55 0.73 NA NA
#> 105 0.65 0.05 0.65 0.55 0.73 NA NA
#> honda_toyota :
#>
#> period mean sd mdn 2.5% 97.5% n_eff Rhat
#> 101 0.73 0.04 0.74 0.65 0.8 NA NA
#> 102 0.73 0.04 0.74 0.65 0.8 NA NA
#> 103 0.73 0.04 0.74 0.65 0.8 NA NA
#> 104 0.73 0.04 0.74 0.65 0.8 NA NA
#> 105 0.73 0.04 0.74 0.65 0.8 NA NA
#> honda_nissan :
#>
#> period mean sd mdn 2.5% 97.5% n_eff Rhat
#> 101 0.64 0.05 0.64 0.54 0.72 NA NA
#> 102 0.64 0.05 0.64 0.54 0.72 NA NA
#> 103 0.64 0.05 0.64 0.54 0.72 NA NA
#> 104 0.64 0.05 0.64 0.54 0.72 NA NA
#> 105 0.64 0.05 0.64 0.54 0.72 NA NAPlot the weighted forecast. Save plots into a ggplot object and post-process
plt <- plot(w_fc, askNewPage = FALSE, type = 'var' )
library( patchwork )
( plt$honda + ggplot2::coord_cartesian(ylim = c(0, 2.5 ) ) ) /
( plt$toyota + ggplot2::coord_cartesian(ylim = c(0, 2.5 ) ) ) /
( plt$nissan + ggplot2::coord_cartesian(ylim = c(0, 2.5 ) ) )
#> Coordinate system already present. Adding new coordinate system, which will replace the existing one.
#> Coordinate system already present. Adding new coordinate system, which will replace the existing one.
#> Coordinate system already present. Adding new coordinate system, which will replace the existing one.
We can add predictors for the constant variance term, c or C, in the MGARCH model with the option xC = The predictors need to be of the same dimension as the time-series object. For example, with three time-series of lenght 100, the predictor needs to be entered as a 100 by 3 matrix as well.
As an example, we will add nissan as the predictor for C in a bivariate MGARCH:
# Fit CCC(1, 1) with constant on the mean structure.
fitx <- bmgarch(stocks.z[1:100, c("toyota", "honda")],
xC = stocks.z[1:100, c("nissan", "nissan")],
parameterization = "CCC",
iterations = 1000,
P = 2, Q = 2,
distribution = "Student_t",
meanstructure = "constant")
#>
#> CHECKING DATA AND PREPROCESSING FOR MODEL 'CCCMGARCH' NOW.
#>
#> COMPILING MODEL 'CCCMGARCH' NOW.
#>
#> STARTING SAMPLER FOR MODEL 'CCCMGARCH' NOW.The preidictors for C are on a log scale in section Exogenous predictor
summary(fitx)
#> Model: CCC-MGARCH
#> Basic Specification: H_t = D_t R D_t
#> diag(D_t) = sqrt(h_[ii,t]) = c_h + a_h*y^2_[t-1] + b_h*h_[ii, t-1
#>
#> Distribution: Student_t
#> ---
#> Iterations: 1000
#> Chains: 4
#> Date: Wed Sep 2 00:16:12 2020
#> Elapsed time (min): 0.58
#>
#> GARCH(2,2) estimates for conditional variance:
#>
#> mean sd mdn 2.5% 97.5% n_eff Rhat
#> a_h_1,ty 0.09 0.09 0.07 0.00 0.33 1786.48 1
#> a_h_1,hn 0.08 0.08 0.05 0.00 0.29 1964.84 1
#> a_h_2,ty 0.10 0.09 0.07 0.00 0.32 2107.94 1
#> a_h_2,hn 0.13 0.13 0.09 0.00 0.50 1681.02 1
#> b_h_1,ty 0.21 0.16 0.17 0.01 0.58 2000.09 1
#> b_h_1,hn 0.18 0.14 0.14 0.00 0.54 2162.12 1
#> b_h_2,ty 0.26 0.17 0.24 0.01 0.62 1974.25 1
#> b_h_2,hn 0.18 0.16 0.15 0.01 0.58 1567.39 1
#> c_h_var_ty 0.22 0.10 0.20 0.07 0.44 1408.81 1
#> c_h_var_hn 0.40 0.15 0.39 0.14 0.73 1542.60 1
#>
#>
#> Constant correlation (R) coefficients:
#>
#> mean sd mdn 2.5% 97.5% n_eff Rhat
#> R_hn-ty 0.73 0.05 0.73 0.62 0.81 2491.25 1
#>
#>
#> Intercept estimates on the location:
#>
#> mean sd mdn 2.5% 97.5% n_eff Rhat
#> (Intercept)_toyota -0.09 0.08 -0.09 -0.26 0.07 1915.77 1
#> (Intercept)_honda -0.05 0.10 -0.05 -0.25 0.13 1931.59 1
#>
#>
#> Exogenous predictor (beta1 on log scale: c = exp( beta_0 + beta_1*x ):
#>
#> mean sd mdn 2.5% 97.5% n_eff Rhat
#> beta0_ty -1.63 0.48 -1.60 -2.64 -0.81 1272.66 1
#> beta0_hn -1.00 0.43 -0.94 -1.95 -0.32 1244.65 1
#> beta_ty -0.18 0.38 -0.19 -0.92 0.61 2001.59 1
#> beta_hn 0.06 0.32 0.08 -0.62 0.65 1630.81 1
#>
#>
#> Df constant student_t (nu):
#>
#> mean sd mdn 2.5% 97.5% n_eff Rhat
#> 45.92 28.82 39.54 8.94 116.75 2758.71 1.00
#>
#>
#> Log density posterior estimate:
#>
#> mean sd mdn 2.5% 97.5% n_eff Rhat
#> -129.67 4.07 -129.39 -138.46 -122.67 798.37 1.00We can generate forecasts given the known values of the predictor. Note that the dimension of the predictor needs to match the number of timepoints that we predict ahead and the number of variables, 5 by 2, in this example:
fc2x <- forecast(fitx, ahead = 5, xC = stocks.z[101:105, c("nissan", "nissan")])
fc2x
#> ---
#> [Variance] Forecast for 5 ahead:
#>
#> toyota :
#>
#> period mean sd mdn 2.5% 97.5% n_eff Rhat
#> 101 0.46 0.11 0.45 0.27 0.71 1890.91 1
#> 102 0.47 0.16 0.45 0.24 0.82 1777.67 1
#> 103 0.51 0.21 0.47 0.23 1.00 1986.70 1
#> 104 0.53 0.24 0.50 0.23 1.11 1880.80 1
#> 105 0.56 0.25 0.52 0.26 1.13 1941.90 1
#> honda :
#>
#> period mean sd mdn 2.5% 97.5% n_eff Rhat
#> 101 0.77 0.14 0.76 0.52 1.09 2052.65 1
#> 102 0.80 0.25 0.77 0.43 1.33 1949.45 1
#> 103 0.89 0.39 0.83 0.43 1.81 1950.88 1
#> 104 0.90 0.41 0.82 0.43 1.86 2028.12 1
#> 105 0.94 0.76 0.83 0.46 2.08 1969.39 1