rugarch

2 messages · pantea hafezian, Alexios Ghalanos

Wed, Feb 13, 2013 9:41 PM #

Dear all
    I have modified my
    program?according to?your?recommendation?but
    unfortunately I still face same problem. I attached the program, data and
    out put for your consideration.
    Thank you very much and?waiting
    to hera from you.
    With many kind,
    
   
  
  
 


?

--- On?Fri, 2/8/13, alexios ghalanos?<alexios at 4dscape.com>?wrote:

????????????????????????????????????????????????????????????? 

From: alexios ghalanos <alexios at 4dscape.com>

Subject: Re: [R-SIG-Finance] Question

To: "pantea hafezian" <pantea_hafezian at yahoo.com>

Cc: r-sig-finance at r-project.org

Date: Friday, February 8, 2013, 1:19 AM

Please
follow the guidelines and post a reproducible example next time.

1. What is data(Gold), where can we find it?

2. There is no model called "egarch" in the specification. It is
"eGARCH"

You are starting the model with 100 data points using the eGARCH model and the
normal distribution. It is quite likely, that the solution converges to one
which is on the boundary of covariance stationarity. I have replicated this :

########################################

set.seed(10)

X = rnorm(1000)

spec =ugarchspec(variance.model=list(model="eGARCH"),distribution.model
= "norm")

roll = ugarchroll(spec, data = X, n.start = 100,? refit.every = 500,
refit.window = "moving", solver = "solnp", fit.control =
list(),

calculate.VaR = TRUE, VaR.alpha = c(0.01, 0.025, 0.05), keep.coef = TRUE)

as.data.frame(roll)

########################################

See the NaNs in the sigma.

Solutions:

1. Use more more data for the estimation start (e.g. 200).

2. Use a different model (e.g. sGARCH, gjrGARCH etc).

Also, read some of the older posts in this forum on the amount of data to use
and why it is a bad idea to use so little data.

-Alexios

?

?

?

?

On 08/02/2013

01:34, pantea hafezian wrote:

package (rugarch). Actually

(VaR) by using this package

problem is related? to the back

test, the value of K (LRUC)? is

equal to 1 for all models. I

various data but the results

example. It would be appreciate

like this or the problem

from you about that.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://stat.ethz.ch/pipermail/r-sig-finance/attachments/20130213/99e5d647/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Gold-Data.xlsx
Type: application/vnd.openxmlformats-officedocument.spreadsheetml.sheet
Size: 42538 bytes
Desc: not available
URL: <https://stat.ethz.ch/pipermail/r-sig-finance/attachments/20130213/99e5d647/attachment.bin>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Out Put.docx
Type: application/vnd.openxmlformats-officedocument.wordprocessingml.document
Size: 14077 bytes
Desc: not available
URL: <https://stat.ethz.ch/pipermail/r-sig-finance/attachments/20130213/99e5d647/attachment-0001.bin>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: Program.text
URL: <https://stat.ethz.ch/pipermail/r-sig-finance/attachments/20130213/99e5d647/attachment.pl>

Alexios Ghalanos

Thu, Feb 14, 2013 2:25 AM #

Check that your code is sensible and reproducible (it is not).
1. You have provided an excel file (bad form - use csv/txt/native R 
format) of prices but have not shown in your file how you go from that 
to "Gold$Rr.1000" in your code.
2. Your code on the VaR report is on another object, "X.bktest"
"report(X.bktest, type="VaR", VaR.alpha = 0.01, conf.level = 0.95)"
rather than on the object you are estimating "mod".
In fact, running the code you provided on the log returns of your 
dataset yields very reasonable results.
3. You provide your error report in an ms word file (bad form) when you 
could have easily provided us with text output.

In future, please put some more thought and consideration into your 
submitted code/questions if you expect any further help.

Regards,
Alexios

On 14/02/2013 05:41, pantea hafezian wrote:

Dear all

I have modified my program according to your recommendation but
unfortunately I still face same problem. I attached the program, data
and out put for your consideration.

Thank you very much and waiting to hera from you.

With many kind,



--- On *Fri, 2/8/13, alexios ghalanos /<alexios at 4dscape.com>/* wrote:


From: alexios ghalanos <alexios at 4dscape.com>
Subject: Re: [R-SIG-Finance] Question
To: "pantea hafezian" <pantea_hafezian at yahoo.com>
Cc: r-sig-finance at r-project.org
Date: Friday, February 8, 2013, 1:19 AM

Please follow the guidelines and post a reproducible example next time.
1. What is data(Gold), where can we find it?
2. There is no model called "egarch" in the specification. It is "eGARCH"

You are starting the model with 100 data points using the eGARCH model
and the normal distribution. It is quite likely, that the solution
converges to one which is on the boundary of covariance stationarity. I
have replicated this :
########################################
set.seed(10)
X = rnorm(1000)
spec =ugarchspec(variance.model=list(model="eGARCH"),distribution.model
= "norm")
roll = ugarchroll(spec, data = X, n.start = 100,  refit.every = 500,
refit.window = "moving", solver = "solnp", fit.control = list(),
calculate.VaR = TRUE, VaR.alpha = c(0.01, 0.025, 0.05), keep.coef = TRUE)
as.data.frame(roll)
########################################
See the NaNs in the sigma.

Solutions:
1. Use more more data for the estimation start (e.g. 200).
2. Use a different model (e.g. sGARCH, gjrGARCH etc).

Also, read some of the older posts in this forum on the amount of data
to use and why it is a bad idea to use so little data.

-Alexios

On 08/02/2013 01:34, pantea hafezian wrote:

Dear Sir/Ms
Thank you for your valuable and very helpful  package (rugarch). Actually
I am going to compare different GARCH models  (VaR) by using this package
but unfortunately  I face a problem. The  problem is related  to the back
testing method. For instance, in Qupiectest, the value of K (LRUC)  is
amounting to 0 and consequently p-value is  equal to 1 for all models. I
performed this procedure several times with  various data but the results
is same as I mentioned even for your own  example. It would be appreciate
if let me know the result is supposed to be  like this or the problem
rises from my programming.
My program is enclosed to this email.
Thank you again and I look forward hearing  from you about that.
Sincerely Yours,
Pantea Hafezian
PhD candidate in Finance,
University Technology Malaysia (UTM)

_______________________________________________
R-SIG-Finance at r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-sig-finance
-- Subscriber-posting only. If you want to post, subscribe first.
-- Also note that this is not the r-help list where general R questions should go.