epoh老师，您好！
您给的这个链接打不开，您能把这个包传上来吗？
非常感谢！
Firefox  无法建立到 127.0.0.1:10489 服务器的连接。

install.packages("rmgarch", repos="http://R-Forge.R-project.org")

or plot(fit.1)

Make a plot selection (or 0 to exit):

1:   Conditional Mean (vs Realized Returns)
2:   Conditional Sigma (vs Realized Absolute Returns)
3:   Conditional Covariance
4:   Conditional Correlation
5:   EW Portfolio Plot with conditional density VaR limits

Selection: 4
  

这个今天也看到了，谢谢了啊

我现在又碰到一个问题了，也是mgarch包的问题。这里面做go-garch是用的基于独立成分ICA的，那是不是就相当于ICA-GARCH啊？那我要做基于主成分的PCA-GARCH应该怎么办？谢谢啊~

你的PC-GARCH可能是指 Patrick Burns (2005)
  Multivariate GARCH with Only Univariate Estimation
  
那可能要照着他的The PC-GARCH Algorithm
七个步骤做了

Newtons<-function (fun, x, ep=1e-5, it_max=100){
    index<-0; k<-1
    while (k<=it_max){
        x1 <- x; obj <- fun(x);
        x  <- x - solve(obj$J, obj$f);
        norm <- sqrt((x-x1) %*% (x-x1))
        if (norm<ep){
            index<-1; break
        }
        k<-k+1
    }
    obj <- fun(x);
    list(root=x, it=k, index=index, FunVal= obj$f)
}

epoh老师，您好！
       x <- x - solve(obj$J, obj$f);
这一行中，solve是什么意思？
obj$J和obj$f是什么意思？J是Jacobi矩阵，是如何求出Jacobi矩阵的？
在 x <- x - solve(obj$J, obj$f);
x - solve(obj$J, obj$f)为什么要赋给x？
x(k+1)<-x(K)-J(逆）*f(x(K))为什么可以x <- x - solve(obj$J, obj$f)这样写呢？

我再看看

Newtons<-function (fun, x, ep=1e-5, it_max=100){
    index<-0; k<-1
    while (k<=it_max){
        x1 <- x; obj <- fun(x);
        x  <- x - solve(obj$J, obj$f);
        norm <- sqrt((x-x1) %*% (x-x1))
        if (norm<ep){
            index<-1; break
        }
        k<-k+1
    }
    obj <- fun(x);
    list(root=x, it=k, index=index, FunVal= obj$f)
}

moment <-function(p){
f<-c(p[1]*p[2]-M1,p[1]*p[2]-p[1]*p[2]^2-M2)
J<-matrix(c(p[2],p[1],p[2]-p[2]^2,p[1]-2*p[1]*p[2]),nrow=2,byrow=T)#Jacobi
list(f=f,J=J)
  }

#####Binomial distribution B(size,prob)
set.seed(123)
y<-rbinom(1000, 20, 0.5);
n<-length(y)
M1<-mean(y);   #9.978
M2<-(n-1)/n*var(y)   #4.977516
p<-c(10,0.7)  # p[1]=k , p[2]=p

Newtons(moment, p)
$root
[1] 19.9101695  0.5011509
$it
[1] 5
$index
[1] 1
$FunVal
[1]  1.826095e-12 -4.090950e-12

##########
答复你的问题
1.obj$J和obj$f是什么意思？如何求出Jacobi矩阵的？
  obj <- fun(x);这里的fun(x)就是moment(p)
  function moment()就已经定义了
  f<-c(p[1]*p[2]-M1,p[1]*p[2]-p[1]*p[2]^2-M2)  #FunVal
  J<-matrix(c(p[2],p[1],p[2]-p[2]^2,p[1]-2*p[1]*p[2]),nrow=2,byrow=T)#Jacobi

2.solve是什么意思？
  solve Solve a System of Equations
      solves the equation a %*% x = b for x, where b can be either a vector or a matrix.
  详细请看帮助?solve
  
3.x - solve(obj$J, obj$f)为什么要赋给x？
  迭代法:
  k=1,x(start value)=[10,0.7]
         经Newton's method,
         x <- x - solve(obj$J, obj$f);
         x=[18.3035020 , 0.4165549]
  这就形成k=2,x(start value)=[18.3035020 , 0.4165549]
                k=3,x(start value)=[19.089003 , 0.527265]
                ...
              当k=5,x(start value)=[19.9101655 , 0.5011505]
                 x value 是[19.9101695  0.5011509]
                其与[19.9101655 , 0.5011505]的norm(4.05753e-06)
                已经小于ep=1e-5,故程序中止

epoh老师，您好！
     您说的我明白了，非常感谢！
在以下程序中，书上说是递归调用，也就是fun 调用fun1,
我不明白：1、为什么要递归调用;
2如何实现fun调用fun1的？
epoh老师，希望您能给我讲讲！
非常感谢！
area <- function(f, a, b, eps = 1.0e-06, lim = 10) {
   fun1 <- function(f, a, b, fa, fb, a0, eps, lim, fun) {
      d <- (a + b)/2; h <- (b - a)/4; fd <- f(d)
      a1 <- h * (fa + fd); a2 <- h * (fd + fb)
      if(abs(a0 - a1 - a2) < eps || lim == 0)
         return(a1 + a2)
      else {
         return(fun(f, a, d, fa, fd, a1, eps, lim - 1, fun)
             + fun(f, d, b, fd, fb, a2, eps, lim - 1, fun))
      }
   }
   fa <- f(a); fb <- f(b); a0 <- ((fa + fb) * (b - a))/2
   fun1(f, a, b, fa, fb, a0, eps, lim, fun1)
}

f <- function(x) 1/x
quad<-area(f,1,5); quad

另外，epoh老师，以下错误如何纠正？
>> A=['first';'second']
??? Error using ==> vertcat
All rows in the bracketed expression must have the same
number of columns.

>>

这个不太好解释
也许你看R function area() Adaptive Numerical Integration,会更清楚
Integrate a function of one variable over a finite range using a recursive adaptive method.
  
The method divides the interval in two and compares the values given by Simpson's rule and the trapezium rule.
If these are within eps of each other the Simpson's rule result is given,
otherwise the process is applied separately to each half of the interval and the results added together.
     http://127.0.0.1:24867/library/MASS/html/area.html

能否把'first','second',列出来以方便解释   

> library(RODBC)
>   z <- odbcConnectExcel('D:/data/mydata.xls')
>  foo<-sqlFetch(z,'Sheet1')
>  close(z)
>   foo
   year gdp ind city  con
1  2001  21  22    22  22
2  2002  22  23    28  24
3  2003  23  24    34  26
4  2004  24  25    40  28
5  2005  25  26    46  30
6  2006  26  27    52  32
7  2007  27  28    58  34
8  2008  28  29    64  36
9  2009  29  30    70  38
10 2010  30  31    76  40
11 2011  31  32    82  42
12 2012  32  33    88  44
13 2013  33  34    94  46
14 2014  34  35   100  48
15 2015  35  36   106  50
16 2016  36  37   112  52
> model1=lm(gdp~ind+city+con)
错误于eval(expr, envir, enclos) : 找不到对象'gdp'
> model1=lm(gdp~ind+city+con,data=mydata)
错误于is.data.frame(data) : 找不到对象'mydata'
> model1=lm(gdp~ind+city+con,data=foo)
错误于eval(expr, envir, enclos) : 找不到对象'city'
> attach(foo)
> model1=lm(gdp~ind+city+con)
错误于eval(expr, envir, enclos) : 找不到对象'city'
> city
错误: 找不到对象'city'
> foo%city
错误: 意外的输入在"foo%city"里
> fix(foo)
> model1=lm(gdp~ind+con,data=foo)
> moedl1
错误: 找不到对象'moedl1'
> model1

Call:
lm(formula = gdp ~ ind + con, data = foo)

Coefficients:
(Intercept)          ind          con  
         -1            1           NA  

>
epoh老师，您好！
以上city 明明读进去了，怎么会提示找不到呢？

library(RODBC)
channel <- odbcConnectExcel("foo.xls")
sqlTables(channel)
#                       TABLE_CAT TABLE_SCHEM TABLE_NAME   TABLE_TYPE REMARKS
# C:\\Users\\...\\Documents\\foo        <NA>       foo$ SYSTEM TABLE    <NA>

foo=sqlFetch(channel, "foo")
foo
#  year        gdp        ind      city         con
#1 2000 0.08431279 0.09429051 0.3622000 0.006607202
#2 2001 0.08300318 0.08441516 0.3766000 0.007220217
#3 2002 0.09082068 0.09829292 0.3908978 0.001303356
#4 2003 0.10025379 0.12671907 0.4053023 0.011064107
#5 2004 0.10085040 0.11112455 0.4176001 0.015448986
#6 2005 0.11310035 0.12082325 0.4299000 0.012044374
#7 2006 0.12676534 0.13391125 0.4434287 0.013780144
#8 2007 0.14162395 0.15063330 0.4588900 0.010812481
#9 2008 0.09634668 0.09877462 0.4698913 0.010299511

model1=lm(gdp~ind+city+con,data=foo)
model1

Call:
lm(formula = gdp ~ ind + city + con, data = foo)

Coefficients:
(Intercept)          ind         city          con  
   -0.03515      0.75016      0.13689     -0.28188  

> library(RODBC)
>  channel <- odbcConnectExcel("foo.xls")
>  sqlTables(channel)
[1] TABLE_CAT   TABLE_SCHEM TABLE_NAME  TABLE_TYPE  REMARKS   
<0 行> (或0-长度的row.names)
>  #                       TABLE_CAT TABLE_SCHEM TABLE_NAME   TABLE_TYPE REMARKS
>  # C:\\Users\\...\\Documents\\foo        <NA>       foo$ SYSTEM TABLE    <NA>
>  
> foo=sqlFetch(channel, "foo")
错误于odbcTableExists(channel, sqtable) :
  ‘foo’: table not found on channel
>  foo
错误: 找不到对象'foo'
>  #  year        gdp        ind      city         con
>  #1 2000 0.08431279 0.09429051 0.3622000 0.006607202
>  #2 2001 0.08300318 0.08441516 0.3766000 0.007220217
>  #3 2002 0.09082068 0.09829292 0.3908978 0.001303356
>  #4 2003 0.10025379 0.12671907 0.4053023 0.011064107
>  #5 2004 0.10085040 0.11112455 0.4176001 0.015448986
>  #6 2005 0.11310035 0.12082325 0.4299000 0.012044374
>  #7 2006 0.12676534 0.13391125 0.4434287 0.013780144
>  #8 2007 0.14162395 0.15063330 0.4588900 0.010812481
>  #9 2008 0.09634668 0.09877462 0.4698913 0.010299511
>  
> model1=lm(gdp~ind+city+con,data=foo)
错误于is.data.frame(data) : 找不到对象'foo'
>  model1
错误: 找不到对象'model1'
> local({pkg <- select.list(sort(.packages(all.available = TRUE)),graphics=TRUE)
+ if(nchar(pkg)) library(pkg, character.only=TRUE)})
>
epoh老师，您好！
为什么在我的电脑上运行会是以上结果？

你有设定ODBC (Open Database Connectivity) for MS-Excel ??

没有设定，不知如何设定？

odbc_settings_excel.pdf
  

??? Error using ==> fmincon
FMINCON cannot continue because user supplied objective function failed with the following error:
Variable 'indep_nS' is used as a command function.
Error in ==> MS_Regress_Fit_tvtp at 67
    [param]=fmincon(@(param)MS_Regress_Lik_tvtp(dep,indep_nS,indep_S,px,param,k,S,advOpt,dispOut),param0, ...
Error in ==> MS_VAR_Fit_tvtp at 68
[Spec_Output]=MS_Regress_Fit_tvtp(dep,indep,px,k,S,advOpt); % Estimating the model
Error in ==> Example_MS_Regress_Fit_MSVAR_tvtp at 21
[Spec_Out]=MS_VAR_Fit_tvtp(dep,nLag,px,k,doIntercept,advOpt);
??? Error using ==> checkInputs_tvtp
So far, in the multivariate version (size(dep,2)>1), the package only handles the multivariate normal distribution. Please set advOpt.distrib='Normal'.
Error in ==> checkInputs_tvtp at 129
        error('So far, in the multivariate version (size(dep,2)>1), the package only handles the multivariate normal distribution. Please set advOpt.distrib=''Normal''.')
Error in ==> MS_Regress_Fit_tvtp at 33
checkInputs_tvtp(); % checking if inputs variables are OK
Error in ==> Example_MS_Regress_Fit_MultiVar_tvtp at 28
[Spec_Out]=MS_Regress_Fit_tvtp(dep,indep,px,k,S,advOpt); % Estimating the model
??? Variable 'indep_nS' is used as a command function.
Error in ==> MS_Regress_Fit_tvtp>@(param)MS_Regress_Lik_tvtp(dep,indep_nS,indep_S,px,param,k,S,advOpt,dispOut) at 81
        [param,fval,exitflag,output]=fminunc(@(param)MS_Regress_Lik_tvtp(dep,indep_nS,indep_S,px,param,k,S,advOpt,dispOut),param0,options);
Error in ==> fminunc at 182
   f = feval(funfcn{3},x,varargin{:});
Error in ==> MS_Regress_Fit_tvtp at 81
        [param,fval,exitflag,output]=fminunc(@(param)MS_Regress_Lik_tvtp(dep,indep_nS,indep_S,px,param,k,S,advOpt,dispOut),param0,options);
Error in ==> Example_MS_Regress_Fit_t_Dist_tvtp at 37
[Spec_Out]=MS_Regress_Fit_tvtp(dep,indep,px,k,S,advOpt); % Estimating the model
??? Error using ==> fmincon
FMINCON cannot continue because user supplied objective function failed with the following error:
Variable 'indep_nS' is used as a command function.
Error in ==> MS_Regress_Fit_tvtp at 67
    [param]=fmincon(@(param)MS_Regress_Lik_tvtp(dep,indep_nS,indep_S,px,param,k,S,advOpt,dispOut),param0, ...
Error in ==> Example_MS_Regress_Fit_tvtp at 37
[Spec_Out]=MS_Regress_Fit_tvtp(dep,indep,px,k,S,advOpt); % Estimating the model
>>
epoh老师，您好！
附件 中的程序和数据为什么运行会出现以上问题？如何解决？

这个程序很容易运行.
除非你是用自己的数据.

>>Example_MS_Regress_Fit_tvtp

***** Numerical Optimization Converged *****

Final log Likelihood: 2494.163
Number of estimated parameters: 10
Type of Switching Model: Univariate
Distribution Assumption -> Normal
Method SE calculation -> 2

***** Final Parameters for Equation #1 *****

---> Non Switching Parameters <---

Non Switching Parameter for Equation #1, Indep column 2
     Value:                0.471093
     Std Error (p. value): 0.026581 (0.0000)
Non Switching Parameter for Equation #1, Indep column 3
     Value:                0.156775
     Std Error (p. value): 0.033011 (0.0000)

--->   Switching Parameters (Distribution Parameters)  <---

State 1
    Model's Variance:      0.000284
    Std Error (p. value):  0.000017 (0.0000)
State 2
    Model's Variance:      0.000893
    Std Error (p. value):  0.000082 (0.0000)

--->   Switching Parameters (Regressors)  <---

Switching Parameters for Equation #1 - Indep column 1

State 1
   Value:                0.000563
   Std Error (p. value): 0.000668 (0.3994)
State 2
   Value:                -0.000130
   Std Error (p. value): 0.002012 (0.9484)

---> Time Varying Transition Probabilities Matrix Estimation (std. error, p-value) <---
      
pa{1,1}
   pa{1,1}(1)
   Value:                9.246796
   Std Error (p. value): 8.332666 (0.2674)
   pa{1,1}(2)
   Value:                176.520977
   Std Error (p. value): 170.157909 (0.2998)
      
pa{1,2}
   pa{1,2}(1)
   Value:                -1.863982
   Std Error (p. value): 0.303879 (0.0000)
   pa{1,2}(2)
   Value:                11.625862
   Std Error (p. value): 14.119926 (0.4105)

---> The Latest Transition Probabilities Matrix <---

      1.0000   0.0407   
      0.0000   0.9593   

---> Expected Duration of Regimes <---

     Expected duration of Regime #1: 59.49 time periods
     Expected duration of Regime #2: 27.50 time periods

  To get started, select MATLAB Help or Demos from the Help menu.

??? Error using ==> fmincon
FMINCON cannot continue because user supplied objective function failed with the following error:
Variable 'indep_nS' is used as a command function.

Error in ==> MS_Regress_Fit_tvtp at 67
    [param]=fmincon(@(param)MS_Regress_Lik_tvtp(dep,indep_nS,indep_S,px,param,k,S,advOpt,dispOut),param0, ...

Error in ==> MS_VAR_Fit_tvtp at 68
[Spec_Output]=MS_Regress_Fit_tvtp(dep,indep,px,k,S,advOpt); % Estimating the model

Error in ==> Example_MS_Regress_Fit_MSVAR_tvtp at 21
[Spec_Out]=MS_VAR_Fit_tvtp(dep,nLag,px,k,doIntercept,advOpt);

>>
epoh老师，您好！
我运行还是出现以上问题，是不是我的fmincon这个matlab 函数有问题，
您能不能把您的fmincon这个函数给我发一份？
非常感谢！

fmincon_R2008a
  

To get started, select MATLAB Help or Demos from the Help menu.
??? Undefined function or variable 'advopt'.
Error in ==> MS_VAR_Fit_tvtp at 68
[Spec_Output]=MS_Regress_Fit_tvtp(dep,indep,px,k,S,advopt); % Estimating the model
Error in ==> Example_MS_Regress_Fit_MSVAR_tvtp at 21
[Spec_Out]=MS_VAR_Fit_tvtp(dep,nLag,px,k,doIntercept,advOpt);
Warning: Name is nonexistent or not a directory: m_Files.
> In path at 115
  In addpath at 95
  In Example_MS_Regress_Fit_MultiVar_tvtp at 5
Warning: Name is nonexistent or not a directory: data_Files.
> In path at 115
  In addpath at 95
  In Example_MS_Regress_Fit_MultiVar_tvtp at 6
??? Error: File: D:\MATLAB7\work\m_Files_tvtp\MS_Regress_Fit_tvtp.m Line: 67 Column: 29
Missing MATLAB operator.
Error in ==> Example_MS_Regress_Fit_MultiVar_tvtp at 28
[Spec_Out]=MS_Regress_Fit_tvtp(dep,indep,px,k,S,advOpt); % Estimating the model
??? Error: File: D:\MATLAB7\work\m_Files_tvtp\MS_Regress_Fit_tvtp.m Line: 67 Column: 29
Missing MATLAB operator.
Error in ==> Example_MS_Regress_Fit_t_Dist_tvtp at 37
[Spec_Out]=MS_Regress_Fit_tvtp(dep,indep,px,k,S,advOpt); % Estimating the model
??? Error: File: D:\MATLAB7\work\m_Files_tvtp\MS_Regress_Fit_tvtp.m Line: 67 Column: 29
Missing MATLAB operator.
Error in ==> Example_MS_Regress_Fit_tvtp at 37
[Spec_Out]=MS_Regress_Fit_tvtp(dep,indep,px,k,S,advOpt); % Estimating the model
>>
epoh老师，您好！
   旧问题解决了，又出现了新问题，您看看能不能解决？

***** Numerical Optimization Converged *****

Final log Likelihood: 2381.2081
Number of estimated parameters: 30
Number of Equations in System: 1
Distribution Assumption -> Normal
Standard error calculation -> 1

***** Final Parameters for Equation #1 *****

     Intercept - Parameter Value (Standard Error, p value)
          State 1, Intercept = 0.00 (0.00,0.00)
          State 2, Intercept = 0.03 (0.00,0.00)
          State 3, Intercept = -0.02 (0.00,0.00)
     Dependent Variable #1 - Parameter Value (Standard Error, p value)
          State 1, Lag 1 = -0.11 (0.00,0.00)
          State 2, Lag 1 = -0.01 (0.00,0.00)
          State 3, Lag 1 = -0.03 (0.00,0.00)
          State 1, Lag 2 = -0.07 (0.00,0.00)
          State 2, Lag 2 = -0.02 (0.00,0.00)
          State 3, Lag 2 = -0.10 (0.00,0.00)


---> Time Varying Transition Probabilities Matrix Estimation (std. error, p-value) <---
      
pa{1,1}
   pa{1,1}(1)
   Value:                3.756518
   Std Error (p. value): 0.000000 (0.0000)
   pa{1,1}(2)
   Value:                100.390312
   Std Error (p. value): 0.000000 (0.0000)
   pa{1,1}(3)
   Value:                62.191360
   Std Error (p. value): 0.000000 (0.0000)
      
pa{1,2}
   pa{1,2}(1)
   Value:                0.711068
   Std Error (p. value): 0.000000 (0.0000)
   pa{1,2}(2)
   Value:                -99.674758
   Std Error (p. value): 0.000000 (0.0000)
   pa{1,2}(3)
   Value:                35.381428
   Std Error (p. value): 0.000000 (0.0000)
      
pa{1,3}
   pa{1,3}(1)
   Value:                -38.680849
   Std Error (p. value): 0.000000 (0.0000)
   pa{1,3}(2)
   Value:                6.752075
   Std Error (p. value): 0.000000 (0.0000)
   pa{1,3}(3)
   Value:                9.486009
   Std Error (p. value): 0.000000 (0.0000)
      
pa{2,1}
   pa{2,1}(1)
   Value:                -63.678697
   Std Error (p. value): 0.000000 (0.0000)
   pa{2,1}(2)
   Value:                6.254026
   Std Error (p. value): 0.000000 (0.0000)
   pa{2,1}(3)
   Value:                4.817729
   Std Error (p. value): 0.000000 (0.0000)
      
pa{2,2}
   pa{2,2}(1)
   Value:                7.488303
   Std Error (p. value): 0.000000 (0.0000)
   pa{2,2}(2)
   Value:                26.651397
   Std Error (p. value): 0.000000 (0.0000)
   pa{2,2}(3)
   Value:                20.439856
   Std Error (p. value): 0.000000 (0.0000)
      
pa{2,3}
   pa{2,3}(1)
   Value:                -1.240812
   Std Error (p. value): 0.000000 (0.0000)
   pa{2,3}(2)
   Value:                40.233594
   Std Error (p. value): 0.000000 (0.0000)
   pa{2,3}(3)
   Value:                26.530402
   Std Error (p. value): 0.000000 (0.0000)

---> Expected Duration of Regimes <---

     Expected duration of Regime #1: 7.10 time periods
     Expected duration of Regime #2: 1.61 time periods
     Expected duration of Regime #3: 4.09 time periods

---> Covariance Matrix <---

State 1
      0.00032 (0.00000,0.00)   
State 2
      0.00059 (0.00000,0.00)   
State 3
      0.00064 (0.00000,0.00)