Ajout de la fonction de GMM itératif

R/covariance.GMM.R

@@ -7,12 +7,13 @@
 
 #' Estimated covariance matrix
 #' @param conditions.vector Vector of moment conditions
-#' @param n Sample size
-#' @param ... Parameters of the vector of moment conditions
+#' @param param Vector of estimated parameters
+#' @param sample Individual data sample
+#' @param ... Functions of the vector of moment conditions
 #' @return A square covariance matrix
 #' 
 #' @author François Pelletier
-covariance.GMM <- function(conditions.vector,n,...)
+covariance.GMM <- function(conditions.vector,param,sample...)
 {
-	t(conditions.vector(...)) %*% conditions.vector(...) / n
+	t(conditions.vector(param,sample,...)) %*% conditions.vector(param,sample,...) / length(sample)
 }

R/iterative.GMM.R

@@ -0,0 +1,39 @@
+# TODO: Add comment
+# 
+# Author: francois
+###############################################################################
+
+
+#' @param start Starting values for the parameters and lagrangian
+#' @param conditions.vector Vector of moment conditions
+#' @param sample Individual data sample
+#' @param ... Functions of the vector of moment conditions
+#' @param W Weighting matrix
+#' @param R Linear constraint matrix of coefficients
+#' @param r Linear constraint constants
+#' @param max.iter Maximum number of iterations
+#' @param epsilon Minimum precision level
+#' @return A list containing the optimized vector of parameter and corresponding covariance matrix
+#' 
+#' @author François Pelletier
+iterative.GMM <- function(start,conditions.vector,sample,...,
+		W,R,r,max.iter=50,epsilon=1E-6)
+{
+	theta1 <- optim.GMM(start,conditions.vector,sample,...,W,R,r)
+	i <- 1
+	repeat
+	{
+		theta2 <- optim.GMM(theta1,conditions.vector,sample,...,W,R,r)
+		S <- covariance.GMM(conditions.vector,param,sample,...)
+		if(sqrt(sum((theta1-theta2)^2))<epsilon)
+			return(list(par=theta2,cov=S))
+		else if (i>max.iter)
+			stop("Iterative GMM does not converge")
+		else
+		{
+			theta1 <- theta2
+			i <- i+1
+		}
+		
+	}
+}

R/objective.GMM.R

@@ -6,19 +6,20 @@
 ###############################################################################
 
 #' Objective function for the GMM method
-#' @param param Vector of parameters to optimize
+#' @param param.lagrangian Vector of parameters and Lagrangian to optimize
 #' @param conditions.vector Vector of moment conditions
-#' @param ... Parameters of the vector of moment conditions
+#' @param sample Individual data sample
+#' @param ... Functions of the vector of moment conditions
 #' @param W Weighting matrix
 #' @param R Linear constraint matrix of coefficients
 #' @param r Linear constraint constants
 #' @return A scalar value
 #' 
 #' @author François Pelletier
-objective.GMM <- function(param.lagrangian,conditions.vector,num.param,...,
+objective.GMM <- function(param.lagrangian,conditions.vector,sample,...,
 		W=diag(length(conditions.vector)),R=0,r=0)
 {
 	param <- param.lagrangian[1:num.param]
 	lagrangian <- param.lagrangian[num.param+1:length(param.lagrangian)]
-	colMeans(conditions.vector(param,...)) %*% ginv(W) %*% colMeans(conditions.vector(param,...))+ abs(t(R %*% param - r) %*% lagrangian)
+	colMeans(conditions.vector(param,sample,...)) %*% ginv(W) %*% colMeans(conditions.vector(param,sample,...))+ abs(t(R %*% param - r) %*% lagrangian)
 }

R/optim.GMM.R

@@ -2,14 +2,15 @@
 #'
 #' @param start Starting values for the parameters and lagrangian
 #' @param conditions.vector Vector of moment conditions
-#' @param number of parameters of the distribution
-#' @param ... Parameters of the vector of moment conditions
+#' @param sample Individual data sample
+#' @param ... Functions of the vector of moment conditions
 #' @param W Weighting matrix
 #' @param R Linear constraint matrix of coefficients
 #' @param r Linear constraint constants
 #' @return une liste contenant le résultat de l'optimisation
+#' 
 #' @author François Pelletier
-optim.GMM <- function(start,conditions.vector,num.param,...,W,R,r)
+optim.GMM <- function(start,conditions.vector,sample,...,W,R,r)
 {
-	optim(c(start,objective.GMM,conditions.vector,num.param,...,W,R,r))
+	optim(start,objective.GMM,conditions.vector,sample,...,W,R,r)
 }

man/covariance.GMM.Rd

@@ -2,12 +2,14 @@
 \alias{covariance.GMM}
 \title{Estimated covariance matrix}
 \usage{
-covariance.GMM(conditions.vector, n, ...)
+covariance.GMM(conditions.vector, param, sample...)
 }
 \arguments{
   \item{conditions.vector}{Vector of moment conditions}
 
-  \item{n}{Sample size}
+  \item{param}{Vector of estimated parameters}
+
+  \item{sample}{Individual data sample}
 
   \item{...}{Parameters of the vector of moment conditions}
 }

man/iterative.GMM.Rd

@@ -0,0 +1,34 @@
+\name{iterative.GMM}
+\alias{iterative.GMM}
+\usage{
+iterative.GMM(start, conditions.vector, sample, ..., W, R, r, max.iter = 50,
+  epsilon = 1e-06)
+}
+\arguments{
+  \item{start}{Starting values for the parameters and
+  lagrangian}
+
+  \item{conditions.vector}{Vector of moment conditions}
+
+  \item{sample}{Individual data sample}
+
+  \item{...}{Functions of the vector of moment conditions}
+
+  \item{W}{Weighting matrix}
+
+  \item{R}{Linear constraint matrix of coefficients}
+
+  \item{r}{Linear constraint constants}
+
+  \item{max.iter}{Maximum number of iterations}
+
+  \item{epsilon}{Minimum precision level}
+}
+\value{
+A list containing the optimized vector of parameter and
+corresponding covariance matrix
+}
+\author{
+François Pelletier
+}
+

man/objective.GMM.Rd

@@ -2,15 +2,18 @@
 \alias{objective.GMM}
 \title{Objective function for the GMM method}
 \usage{
-objective.GMM(param.lagrangian, conditions.vector, num.param, ...,
+objective.GMM(param.lagrangian, conditions.vector, sample, ...,
   W = diag(length(conditions.vector)), R = 0, r = 0)
 }
 \arguments{
-  \item{param}{Vector of parameters to optimize}
+  \item{param.lagrangian}{Vector of parameters and
+  Lagrangian to optimize}
 
   \item{conditions.vector}{Vector of moment conditions}
 
-  \item{...}{Parameters of the vector of moment conditions}
+  \item{sample}{Individual data sample}
+
+  \item{...}{Functions of the vector of moment conditions}
 
   \item{W}{Weighting matrix}
 

man/optim.GMM.Rd

@@ -2,7 +2,7 @@
 \alias{optim.GMM}
 \title{Optimization with constraints for GMM methos}
 \usage{
-optim.GMM(start, conditions.vector, num.param, ..., W, R, r)
+optim.GMM(start, conditions.vector, sample, ..., W, R, r)
 }
 \arguments{
   \item{start}{Starting values for the parameters and
@@ -10,9 +10,9 @@ optim.GMM(start, conditions.vector, num.param, ..., W, R, r)
 
   \item{conditions.vector}{Vector of moment conditions}
 
-  \item{number}{of parameters of the distribution}
+  \item{sample}{Individual data sample}
 
-  \item{...}{Parameters of the vector of moment conditions}
+  \item{...}{Functions of the vector of moment conditions}
 
   \item{W}{Weighting matrix}