Constantine Tsardounis
2005-Dec-25 22:25 UTC
[R] Different ARCH results in R and Eviews using garch from tseries
Dear Sir,
First of all Happy Holidays!,...
I am writing to you because I am a bit confused about ARCH estimation.
Is there a way to find what garch() exactly does, without the need of
reading the source code (because I cannot understand it)?
In Eviews (the results at the end) I am getting different results than
in R (for those that have the program I do: Quick -> Estimage Equation
-> Method: ARCH -> y c x -> GARCH:0 & ARCH:1 -> ARCH-M term:
none.
Data can be downloaded from
http://constantine.evangelopoulos.com/1.2.2-askhseis.econometrix.csv
and can be loaded in R with:
x <- ts(read.csv("1.2.2-askhseis.econometrix.csv")[ ,1])
y <- ts(read.csv("1.2.2-askhseis.econometrix.csv")[ ,2])
garch(summary(lm(y ~ x))$resid^2, c(0,1))
What I am doing wrong? Because I want to check for ARCH(q) effect and
then estimate the final equations (Y on X, with the equation of the
error term)
Thank very much in advance for your assistance,
Tsardounis Constantine
Student in Economics at University of Thessaly, Greece
Eviews results:
Dependent Variable: Y
Method: ML - ARCH
Date: 12/26/05 Time: 00:05
Sample(adjusted): 1 83
Included observations: 83 after adjusting endpoints
Convergence achieved after 16 iterations
Coefficient Std. Error z-Statistic Prob.
C 0.005268 0.002442 2.157327 0.0310
X 0.947425 0.024682 38.38587 0.0000
Variance Equation
C 0.000456 8.55E-05 5.333923 0.0000
ARCH(1) -0.041617 0.117458 -0.354311 0.7231
R-squared 0.941163 Mean dependent var 0.016895
Adjusted R-squared 0.938928 S.D. dependent var 0.086783
S.E. of regression 0.021446 Akaike info criterion -4.801068
Sum squared resid 0.036336 Schwarz criterion -4.684498
Log likelihood 203.2443 F-statistic 421.2279
Durbin-Watson stat 1.503765 Prob(F-statistic) 0.000000
Spencer Graves
2005-Dec-30 18:15 UTC
[R] Different ARCH results in R and Eviews using garch from tseries
Have you tried the garch modeling in the fSeries package? Also, have you tried to think of an example so small and simple you can work it yourself either entirely by hand or using something more transparent? For example, I often use the "Solver" in Excel to minimize a log(likelihood). When I can work a problem that way and then get the same answer from R code, it increases my confidence that I know what R is doing. (If you have Excel but have never used the Solver, you may need to look first at Tools -> Add-Ins -> Solver. Then "Tools -> Solver" should give it to you.) Or write a function to compute the negative of the log(likelihood) and use optim to minimize it, with hessian = TRUE to get the observed information, whose inverse is the variance for the Wald approximation. hope this helps. spencer graves Constantine Tsardounis wrote:> Dear Sir, > > First of all Happy Holidays!,... > > I am writing to you because I am a bit confused about ARCH estimation. > Is there a way to find what garch() exactly does, without the need of > reading the source code (because I cannot understand it)? > In Eviews (the results at the end) I am getting different results than > in R (for those that have the program I do: Quick -> Estimage Equation > -> Method: ARCH -> y c x -> GARCH:0 & ARCH:1 -> ARCH-M term: none. > > Data can be downloaded from > http://constantine.evangelopoulos.com/1.2.2-askhseis.econometrix.csv > and can be loaded in R with: > > x <- ts(read.csv("1.2.2-askhseis.econometrix.csv")[ ,1]) > y <- ts(read.csv("1.2.2-askhseis.econometrix.csv")[ ,2]) > garch(summary(lm(y ~ x))$resid^2, c(0,1)) > > What I am doing wrong? Because I want to check for ARCH(q) effect and > then estimate the final equations (Y on X, with the equation of the > error term) > > > > Thank very much in advance for your assistance, > > Tsardounis Constantine > Student in Economics at University of Thessaly, Greece > > > Eviews results: > Dependent Variable: Y > Method: ML - ARCH > Date: 12/26/05 Time: 00:05 > Sample(adjusted): 1 83 > Included observations: 83 after adjusting endpoints > Convergence achieved after 16 iterations > > Coefficient Std. Error z-Statistic Prob. > > C 0.005268 0.002442 2.157327 0.0310 > X 0.947425 0.024682 38.38587 0.0000 > > Variance Equation > > C 0.000456 8.55E-05 5.333923 0.0000 > ARCH(1) -0.041617 0.117458 -0.354311 0.7231 > > R-squared 0.941163 Mean dependent var 0.016895 > Adjusted R-squared 0.938928 S.D. dependent var 0.086783 > S.E. of regression 0.021446 Akaike info criterion -4.801068 > Sum squared resid 0.036336 Schwarz criterion -4.684498 > Log likelihood 203.2443 F-statistic 421.2279 > Durbin-Watson stat 1.503765 Prob(F-statistic) 0.000000 > > ______________________________________________ > R-help at stat.math.ethz.ch mailing list > https://stat.ethz.ch/mailman/listinfo/r-help > PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html-- Spencer Graves, PhD Senior Development Engineer PDF Solutions, Inc. 333 West San Carlos Street Suite 700 San Jose, CA 95110, USA spencer.graves at pdf.com www.pdf.com <http://www.pdf.com> Tel: 408-938-4420 Fax: 408-280-7915