Skip to content

Asymmetry parameter misspecification in EGARCH model using the rugarch package

3 messages · Joakim Lindboe Brüchmann ., Diego Peroni, Alexios Ghalanos

Joakim Lindboe Brüchmann .
# 
Dear Alexios,

First of all thanks for a wonderful package that I frequently use.

I have estimated an ARMA(1,1)-EGARCH(1) model, and was puzzled by the
outcome for a very long time since I got an insignificant estimate for the
|z_t-1| term (regular ARCH effects) while the asymmetric parameter in front
of z_t-1 was indeed statistically significant. My data looks like normal
financial data.

I based this on the specifications using the ?ugarchspec function in my
script where it says:

*Variance Model (GJR, EGARCH)*
- assymetry term: gamma1

By drawing the news impact curves and reviewing my code and plots, I came
to the conclusion that there was no asymmetry apparent from the data but
that there indeed should bed regular ARCH effects. By running the code for
gjrGARCH also, I noticed that this model confirmed my hypothesis.

Thereby my question:
Isn't there an error in the instructions? Doesn't gamma1 represent regular
ARCH effects while alpha1 is the correct leverage parameter representing
asymmetry?

Best Regards,
Joakim
Diego Peroni
# 
Hi All,

my strategy exit an open position with a simple Signal based on 1 minute 
OHLC bars.

This is my exit rule:

add.rule(qs.strategy, name='ruleSignal',
          arguments=list(sigcol='upTrend', sigval=TRUE,
                         replace=TRUE,
                         orderside='short',
                         ordertype='market',
                         orderqty='all',
                         orderset='ocoshort'
          ),
          type='exit',
          label='ExitShort',
          enabled=TRUE)

My question is: which price quantstrat use to sell market by default?
- Current Bar Close Price?
- Next Bar Open Price?
- Next Bar "worst"/"best" Price?

Thanks in advance

Diego
Alexios Ghalanos
# 
See the vignette section 2.2.3 equation 14. There is no confusion as to why gamma is the asymmetry effect (or more correctly the magnitude effect), whereas alpha is the sign effect (as the vignette states).

You can also check ugarchbench("published") and the underlying code for more details.

Alexios

Sent from my iPhone