热卖商品
新闻详情
关于多元GARCH模型的几种编程例子 - R语言论坛..._人大经济论坛
来自 : 经管之家
发布时间:2021-03-24
More on Multivariate GARCH
We’ll now discuss adjustments to the basic multivariate GARCH models and
other types of analysis which can be done with results from the GARCH instruction.
We’ll look at two new data sets. For Sections 6.1 and 6.2 we will use (a reconstruction
of) the data set from Hafner and Herwartz (2006) (from now on
HH). The full data set (called HHDATA.XLS) has 3270 daily observations on ten
exchange rates vs the dollar, running from 31 December 1979 to 1 April 1994.
The data as used in the paper are expressed in local currency/USD, while the
data on the file are USD/local currency, so we have to change the sign when
computing the returns.1 There’s a separate date column, with the date coded
numerically as a six digit number yymmdd. Although the data set includes all
five days a week and so could be handled as CALENDAR(D) with RATS, we’ll
treat it as irregular, and show how to locate an entry based upon a coded date
field like that. The data are read with
open data hhdata.xls
data(format=xls,org=columns) 1 3720 usxjpn usxfra usxsui $
usxnld usxuk usxbel usxger usxswe usxcan usxita date
Our focus will be on bivariate models on returns for two of the currencies:
the British pound and the Deutsche mark. The authors choose to use separate
univariate autoregressions for the mean models—if you estimate a one lag VAR,
the “other” lags have t-stats less than 1, so leaving them out isn’t unreasonable.
The mean model can be set up with
set demret = -100.0*log(usxger/usxger{1})
set gbpret = -100.0*log(usxuk/usxuk{1})
*
equation demeqn demret
# constant demret{1}
equation gbpeqn gbpret
# constant gbpret{1}
group uniar1 demeqn gbpeqn
The GROUP instruction combines the separate equations into a single model for
input into GARCH. This is the only way to create a mean model with different
http://www.estima.com/ecourse/samples/GARCHSampleChapter.pdf
We’ll now discuss adjustments to the basic multivariate GARCH models and
other types of analysis which can be done with results from the GARCH instruction.
We’ll look at two new data sets. For Sections 6.1 and 6.2 we will use (a reconstruction
of) the data set from Hafner and Herwartz (2006) (from now on
HH). The full data set (called HHDATA.XLS) has 3270 daily observations on ten
exchange rates vs the dollar, running from 31 December 1979 to 1 April 1994.
The data as used in the paper are expressed in local currency/USD, while the
data on the file are USD/local currency, so we have to change the sign when
computing the returns.1 There’s a separate date column, with the date coded
numerically as a six digit number yymmdd. Although the data set includes all
five days a week and so could be handled as CALENDAR(D) with RATS, we’ll
treat it as irregular, and show how to locate an entry based upon a coded date
field like that. The data are read with
open data hhdata.xls
data(format=xls,org=columns) 1 3720 usxjpn usxfra usxsui $
usxnld usxuk usxbel usxger usxswe usxcan usxita date
Our focus will be on bivariate models on returns for two of the currencies:
the British pound and the Deutsche mark. The authors choose to use separate
univariate autoregressions for the mean models—if you estimate a one lag VAR,
the “other” lags have t-stats less than 1, so leaving them out isn’t unreasonable.
The mean model can be set up with
set demret = -100.0*log(usxger/usxger{1})
set gbpret = -100.0*log(usxuk/usxuk{1})
*
equation demeqn demret
# constant demret{1}
equation gbpeqn gbpret
# constant gbpret{1}
group uniar1 demeqn gbpeqn
The GROUP instruction combines the separate equations into a single model for
input into GARCH. This is the only way to create a mean model with different
http://www.estima.com/ecourse/samples/GARCHSampleChapter.pdf
本文链接: http://gachgroup.immuno-online.com/view-685900.html
发布于 : 2021-03-24
阅读(0)
最新动态
2021-03-24
2021-03-24
2021-03-24
2021-03-24
2021-03-24
2021-03-24
2021-03-24
2021-03-24
2021-03-24
2021-03-24
2021-03-24
2021-03-24
联络我们