library(timeSeries)
library(MTS)
library(Rblpapi)

blpConnect(host='192.168.9.61')

data <- feather::read_feather("/home/serenitas/CorpCDOs/data/index_returns.fth")
data$date <- as.Date(data$date)
df <- bdh(paste("VIX", "Index"), "PX_LAST", start.date=as.Date("2009-03-20"))
df <- as.tibble(df)

R <- na.omit(returns)
R <- scale(R, scale=F)
chol <- MCholV(na.omit(returns))

ema <- function(x, alpha=0.1, init=x[1]){
    ## exponential moving average with parameter lambda=1-beta
    filter(alpha*x, filter = 1-alpha, method = "recursive", init = init)
}

ema.slow <- function(y, lambda, init=y[1]) {
    # slower but more explicit
    mu <- init
    mu <- vapply(y, function(x) mu <<- mu*lambda + x*(1 - lambda), numeric(1))
    return( mu )
}

cov.ewm <- function(X, lambda, init=X[1,]) {
    ema(X
beta.ewm <- function(R, lambda, span) {
    # computes beta between two assets using exponential moving averages
    if(ncol(R) != 2) {
        stop("only works for two assets")
    }
    if(missing(lambda)) {
        alpha <- 2/(span+1)
    } else {
        alpha <- 1-lambda
    }
    R <- scale(R, scale=F)
    cov12 <- ema(R[,1] * R[,2], alpha)
    var1 <- ema(R[,1] * R[,1], alpha)
    return ( cov12/var1 )
}

## library(rugarch)
## spec <- ugarchspec(variance.model=list(model="sGARCH"),
##                    mean.model = list(armaOrder=c(0,0), include.mean = TRUE))
## fit.ig <- ugarchfit(spec, na.omit(df$ig), solver='hybrid')
## fit.hy <- ugarchfit(spec, na.omit(df$hy), solver='hybrid')