s-news
[Top] [All Lists]

[S] GARCH(1,1) model

To: s-news@wubios.wustl.edu
Subject: [S] GARCH(1,1) model
From: "Hoon Kim" <hoon@andrew.cmu.edu>
Date: Sat, 24 Apr 1999 13:49:53 -0400
Cc: hoon@andrew.cmu.edu
Sender: owner-s-news@wubios.wustl.edu
Dear S-plus users

I am running a GARCH(1,1) model on S-plus version 3.4 for Sun SPARC and
GARCH module. I tried to run garch model on quarterly ROE(return on equity)
of a firm. The number of observation is 80 (20 years*4 quarters per year). 

The inputs are (converted using rts)

> PetRoe$V5
 1:  0.02396  0.01904  0.02510  0.03949  0.03327  0.04647  0.05763  0.07152
 9:  0.05111  0.06074  0.06614  0.06714  0.05407  0.06087  0.05317  0.04835
17:  0.04126  0.04950  0.04996  0.04304  0.03058  0.04266  0.04097  0.02425
25:  0.02623  0.02323  0.02477  0.04254  0.02005  0.02342  0.01829  0.01683
33:  0.01083  0.00263 -0.00053 -0.26746  0.00394  0.00905  0.00982 -0.08220
41: -0.00871  0.01296  0.01399  0.03370  0.00853  0.02242  0.00331  0.02716
49:  0.02287  0.03472  0.04367  0.02043  0.03363 -0.09143  0.03290  0.02171
57: -0.00089  0.02265  0.01218  0.01806  0.01956  0.01856  0.01653  0.01643
65:  0.01941  0.02215  0.02909  0.01429  0.01250  0.01021  0.00481 -0.14832
73:  0.01834  0.02202  0.03508  0.03926  0.02969  0.02626  0.03957  0.02956
 start deltat frequency 
     1      1         1


The outputs are :

> PetRoe$V5.garch_garch(PetRoe$V5~-1, ~garch(1,1))
Iteration   0  Step Size =  1.00000  Likelihood = -1.00000e+10 

.....

Iteration  41  Step Size = 1.28000e-05  Likelihood =  131.197 
Iteration  41  Step Size = 2.56000e-06  Likelihood =  131.197 
Iteration  41  Step Size = 5.12000e-07  Likelihood =  131.197 
Iteration  41  Step Size = 1.02400e-07  Likelihood =  131.197 

Could not find increasing step size

> summary(PetRoe$V5.garch)

Call: garch(formula.mean = PetRoe$V5 ~ -1, formula.var =  ~ garch(1, 1))

Mean Equation: PetRoe$V5 ~ -1

Conditional Variance Equation:  ~ garch(1, 1)

Conditional Distribution:  gaussian

--------------------------------------------------------------

Estimated Coefficients:
--------------------------------------------------------------
              Value Std.Error t value  Pr(>|t|) 
       A  0.0004075 0.0003919   1.040 1.509e-01
 ARCH(1) -0.0300954 0.0202477  -1.486 7.063e-02
GARCH(1)  0.8172826 0.1861506   4.390 1.779e-05

--------------------------------------------------------------

AIC(3) = -256.3941
BIC(3) = -249.248

Normality Test:
--------------------------------------------------------------
 Jarque-Bera P-value Shapiro-Wilk P-value 
        1212       0       0.6172       0

Ljung-Box test for standardized residuals:
--------------------------------------------------------------
 Statistic P-value Chi^2-d.f. 
     21.08 0.04928         12

Ljung-Box test for squared standardized residuals:
--------------------------------------------------------------
 Statistic P-value Chi^2-d.f. 
     3.694  0.9883         12

Based on the Result, the garch and arch coefficients look significant. But I
guess that is because there was something wrong on computations of
convergence. Is there any way to fix this problem? I tried to increase
n.iter and changed to stepsize=0, but the results are the same. I need some
help and any suggestions are all welcome.

Thanks a million to read this long message. 


Hoon Kim
Graduate School of Industrial Administration
Carnegie Mellon University

E-Mail: hoon@andrew.cmu.edu
URL: www.gsia.cmu.edu/andrew/hoon/www/

-----------------------------------------------------------------------
This message was distributed by s-news@wubios.wustl.edu.  To unsubscribe
send e-mail to s-news-request@wubios.wustl.edu with the BODY of the
message:  unsubscribe s-news

<Prev in Thread] Current Thread [Next in Thread>
  • [S] GARCH(1,1) model, Hoon Kim <=