An embedded and charset-unspecified text was scrubbed... Name: not available URL: <https://stat.ethz.ch/pipermail/r-help/attachments/20130312/b0f70649/attachment.pl>
rugarch: GARCH with Johnson Su innovations
2 messages · Wyss Patrick, Patrick Burns
You want to give returns rather than prices to the garch fitting function. Log returns are more appropriate than simple returns. Actually a negative lambda is what I would expect. Higher volatility (across time) is usually associated with lower returns. The risk premium is more likely a cross-sectional phenomenon than a time series one. Some people are not so believing in risk premia in the first place -- see for instance, Eric Falkenstein. This would have been better sent to the r-sig-finance list. Pat
On 12/03/2013 12:15, Wyss Patrick wrote:
Hey,
I'm trying to implement a GARCH model with Johnson-Su innovations in order to simulate returns of financial asset. The model should look like this:
r_t = alpha + lambda*sqrt(h_t) + sqrt(h_t)*epsilon_t
h_t = alpha0 + alpha1*epsilon_(t-1)^2 + beta1 * h_(t-1).
Alpha refers to a risk-free return, lambda to the risk-premium.
I've implemented it like this:
#specification of the model
spec = ugarchspec(variance.model = list(model = "sGARCH",
garchOrder = c(1,1), submodel = NULL, external.regressors =
NULL, variance.targeting = FALSE), mean.model = list(
armaOrder = c(0,0), include.mean = TRUE, archm = TRUE, archpow = 1,
arfima = FALSE, external.regressors = NULL, archex = FALSE),
distribution.model = "jsu", start.pars = list(), fixed.pars = list())
#fit the model to historical closing price (prices)
fit = ugarchfit(data = prices, spec = spec)
#save coefficients of the fitted model into 'par'
par <- coef(fit)
m = coef(fit)["mu"]
lambda = coef(fit)["archm"]
gamma = coef(fit)["skew"]
delta = coef(fit)["shape"]
#GARCH parameter
a0 = coef(fit)["omega"]
a1 = coef(fit)["alpha1"]
b1 = coef(fit)["beta1"]
My problem is that I often get negative values for lambda, i.e. for the intended risk-premium. So I'm wondering if I've made a mistake in the implementation, as one would usually expect a positive lambda.
And a second question is about the Johnson-Su distribution: Am I right by extracting the Johnson-Su parameters gamma (delta) by the keywords "skew" ("shape")?
Many thanks in advance,
Patrick
[[alternative HTML version deleted]]
______________________________________________ R-help at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Patrick Burns pburns at pburns.seanet.com twitter: @burnsstat @portfolioprobe http://www.portfolioprobe.com/blog http://www.burns-stat.com (home of: 'Impatient R' 'The R Inferno' 'Tao Te Programming')