[R-SIG-Finance] Rugarch Package - Problem with Arch-lm-test

[alexios]

I would have thought that is exactly what you want to happen once you filter
the data for garch effects...that none are left. The ARCH LM test under the
Null of no ARCH effects, as presented in the summary, is applied to the
standardized residuals (i.e. the residuals from the mean filtration
standardized by the conditional standard deviation from the variance
equation). This is clearly explained in both the documentation and the
vignette.

Regards,
Alexios


barb wrote
> 
> Hey guys,
> 
> got some strange results using the rugarch, but can´t figure out what is
> going wrong.
> 
> 	library(rugarch)
> 	library(quantmod)
> 	getSymbols("ADS.DE", from ="2006-01-01", to="2011-12-02")
> 	oc<-ADS.DE[,1]
> 	ccr<- function(x) {100*(log((x[-1])/(x[1:length(x)-1])))} #Continously
> Compounded Returns
> 	GOOGCCR<-apply(oc,2,ccr) # Apply for Matrix-data
> 	spec = ugarchspec(variance.model = list(model = "sGARCH", garchOrder =
> c(1,1)))
> 	fit = ugarchfit(data = GOOGCCR, spec = spec)
> 	fit
> 
> So load the Adidas rates. The time horizon doesn´t really matter for me,
> but i need at least
> 100 data points for the package.
> 
> Optimal Parameters
> ------------------------------------
>         Estimate  Std. Error  t value Pr(>|t|)
> mu      0.051503    0.037037   1.3906 0.164359
> ar1     0.513259    0.217548   2.3593 0.018310
> ma1    -0.565843    0.214081  -2.6431 0.008215
> omega   0.170118    0.046981   3.6210 0.000293
> alpha1  0.115982    0.021801   5.3201 0.000000
> beta1   0.847845    0.027334  31.0176 0.000000
> 
> Information Criteria
> ------------------------------------
>                    
> Akaike       4.1369
> Bayes        4.1579
> Shibata      4.1369
> Hannan-Quinn 4.1447
> 
> Q-Statistics on Standardized Residuals
> ------------------------------------
>       statistic p-value
> Lag10     1.687  0.9891
> Lag15     4.983  0.9756
> Lag20     9.918  0.9345
> 
> H0 : No serial correlation
> 
> Q-Statistics on Standardized Squared Residuals
> ------------------------------------
>       statistic p-value
> Lag10     3.229  0.9192
> Lag15     4.701  0.9812
> Lag20     5.305  0.9983
> 
> ARCH LM Tests
> ------------------------------------
>              Statistic DoF P-Value
> ARCH Lag[2]     0.5589   2  0.7562
> ARCH Lag[5]     1.2106   5  0.9439
> ARCH Lag[10]    3.2138  10  0.9759
> 
> 
> So, basicly i can not reject the null hypothese, that there are no arch
> effects. 
> What is going wrong?
> 


--
View this message in context: http://r.789695.n4.nabble.com/Rugarch-Package-Problem-with-Arch-lm-test-tp4207079p4212293.html
Sent from the Rmetrics mailing list archive at Nabble.com.