[R] Use fitted Garch models in linear regression

[henmarco]

Hi
 I am analysing a data set of daily S&P 500 Index returns and my goal is to
elaborate a relationship with a sentiment indicator (daily data). 
For this purpose I fitted a model to each variable. I found that a GARCH
(1,1) suits best for the differenced closing price of the SPX and a GARCH
(2,2) for the SPX returns.

The sentiment indicator follows a ARMA (2,2) process.

But now I am stuck. How do I use these fitted models to perform a linear
regression on the variables?

Without correction A Model like

model=lm(spxclose-spxsentiment) is in my mind. But this simple method does
not work with garch objects.

The only two alternatives I tried were:

A.one: Find the relationships by evaluating the cross correllograms:

par(mfrow=c(2,2))
both<-ts.union(garchdspxclose$resid,arimaspxpcr$resid)
acf(both,na.action = na.pass) 
pacf(both,na.action = na.pass)

A.two: A paper mentions to correct with NeweyWest for autocorrelation and
heteroskedasticity

  result <- dynlm(spxclose ~ lag(spxclose,1) +lag(spxpcr,1)+lag(vixpcr,1))
  NeweyWest(result)
  coeftest(result, vcov = NeweyWest)

Is this method also correcting for ARCH effects?


Are VAR-modells or the cointegration from Granger and Engle appropriate
tools to analysis daily exchange data comparing returns and sentiment
indicators?

Thank you for your help

marco

--
View this message in context: http://r.789695.n4.nabble.com/Use-fitted-Garch-models-in-linear-regression-tp4632648.html
Sent from the R help mailing list archive at Nabble.com.