R-CODE
setwd("~/Sinha")
# Library locations #
library("MTS", lib.loc="~/R/win-library/3.3")
library("MSBVAR", lib.loc="~/R/win-library/3.3")
library("bvarsv", lib.loc="~/R/win-library/3.3")
library("urca", lib.loc="~/R/win-library/3.3")
library("uroot", lib.loc="~/R/win-library/3.3")
library("vars", lib.loc="~/R/win-library/3.3")
library("tsDyn", lib.loc="~/R/win-library/3.3")
library("TSA", lib.loc="~/R/win-library/3.3")
# Data generating Process for Baysean VAR in csv #
Gendata1 = read.csv("BVARDAT.csv", header = True)
View(Gendata1)
gdp = Gendata1[,2]
cpi = Gendata1[,3]
Yt = ts(gdp, frequency = 4, start = 2010)
Xt = ts(cpi, frequency = 4, start = 2010)
Xt
win.graph(width=7.5, height=4.5,pointsize=12)
par(mfrow=c(1,1))
plot(Yt, ylab="GDP of India", xlab= "Time ", type = "l",
main = "Fig. 4.1 Time plot of quarterly GDP 2010-2017")
plot(Xt, ylab="Consumer's Price Index (CPI)", xlab= "Time", type = "l",
main = "Fig. 4.2 : Time plot of Quarterly Consumers Price Index, 2010-2017")
## IDENTIFYING TRENDS AND SEASONS
##ACF AND PACF
win.graph(width = 8, height = 6.5, pointsize = 12)
par(mfrow=c(2,2))
acf(Yt, lag.max = 72, plot = TRUE, main = "ACF plot of GDP")
pacf(Yt, lag.max = 72, plot = TRUE, main = "PACF plot of GDP")
acf(Xt, lag.max = 56, plot = TRUE, main = "ACF plot of India Consumer Price Index")
pacf(Xt, lag.max = 56, plot = TRUE, main = "PACF plot of India Consumer rice
Index")
x = ur.df(Yt, type = "none", selectlags = "AIC")
xt = ur.df(Yt, type = "trend", selectlags = "AIC")
xd = ur.df(Yt, type = "drift", selectlags = "AIC")
x_1 = ur.df(diff(Yt, lag = 1), type = "none", selectlags = "AIC")
xt_1 = ur.df(diff(Yt, lag = 1), type = "trend", selectlags = "AIC")
xd_1 = ur.df(diff(Yt, lag = 1), type = "drift", selectlags = "AIC")
x_2 = ur.df(diff(diff(Yt), lag = 1), type = "none", selectlags = "AIC")
xt_2 = ur.df(diff(diff(Yt), lag = 1), type = "trend", selectlags = "AIC")
xd_2 = ur.df(diff(diff(Yt), lag = 1), type = "drift", selectlags = "AIC")
summary(xd); summary(xd_1); summary(xd_2)
summary(xt); summary(xt_1); summary(xt_2)
y = ur.df(Xt, type = "none", selectlags = "AIC")
yt = ur.df(Xt, type = "trend", selectlags = "AIC")
yd = ur.df(Xt, type = "drift", selectlags = "AIC")
y_1 = ur.df(diff(Xt, lag = 1), type = "none", selectlags = "AIC")
yt_1 = ur.df(diff(Xt, lag = 1), type = "trend", selectlags = "AIC")
yd_1 = ur.df(diff(Xt, lag = 1), type = "drift", selectlags = "AIC")
y_2 = ur.df(diff(diff(Xt), lag = 1), type = "none", selectlags = "AIC")
yt_2 = ur.df(diff(diff(Xt), lag = 1), type = "trend", selectlags = "AIC")
yd_2 = ur.df(diff(diff(Xt), lag = 1), type = "drift", selectlags = "AIC")
summary(x); x@cval; x@lags; x@teststat
summary(y); y@cval; y@lags; y@teststat
summary(x_1); x_1@cval; x_1@lags; x_1@teststat
summary(y_1); y_1@cval; y_1@lags; y_1@teststat
summary(xt); xt@cval; xt@lags; xt@teststat
summary(yt); yt@cval; yt@lags; yt@teststat
summary(xt_1); xt_1@cval; xt_1@lags; xt_1@teststat
summary(yt_1); yt_1@cval; yt_1@lags; yt_1@teststat
summary(xd); xd@cval; xd@lags; xd@teststat
summary(yd); yd@cval; yd@lags; yd@teststat
summary(xd_1); xd_1@cval; xd_1@lags; xd_1@teststat
summary(yd_1); yd_1@cval; yd_1@lags; yd_1@teststat
### Cointegraion
lg.reg <-lm(Yt~Xt)
error<- lg.reg$residuals
RES = ur.df(error, type = "drift", selectlags = "AIC")
summary(RES)
co.data = cbind(gdp,cpi)
VARselect(co.data, type = "const")$selection
cointest = ca.jo(co.data, 10, type = "trace", ecdet = "const",
spec = "transitory")
cointest
cointest@cval
cointest@teststat[2]
cointest@teststat[1]
gdp = gdp
cpi = cpi
d.gdp = diff(gdp)
d.cpi = diff(cpi)
adf.test(gdp); adf.test(cpi)
ur.df(gdp, type = "drift", selectlags = "AIC")@cval
modelvar = VAR(cbind(gdp,cpi), type = "const", lag.max = 2, ic = "AIC")
modelvar$varresult
summary(modelvar)
data("mts-examples",package="MTS")
z=log(co.data[,1:2])
zt=diffM(z)*100
C=0.1*diag(rep(1,3))
V0=diag(rep(1,2))
model = BVAR(zt,p=1,C,V0)
summary(model)
model$residuals
model$residuals
forecast(model, nsteps = 29)
### MSBVAR forecasts
# Fit model
m1 <- msbvar(Y.sample1, p=1, h=2, lambda0=0.8, lambda1=0.2,
lambda3=1, lambda4=0.2, lambda5=0, mu5=0, mu6=0,
qm=12, prior=0)
# Gibbs sampling
m1id <- gibbs.msbvar(m1, N1=1000, N2=10000, permute=FALSE, Sigma.idx=1)
# Forecast density estimation
msforc <- forecast(m1id, nsteps=nrow(Y.sample2), N1=1000, N2=10000)
# Summarize forecasts
apply(msforc$forecasts, c(2,3), mean)
