EquationsModelling {fMultivar} | R Documentation |
A collection and description of easy to use
functions to perform fits of systems of
regression equations. The underlying functions
are those from the contributed R-package
systemfit written by Jeff D. Hamann and Arne
Henningsen. The package offers functions for
fitting linear structural equations using
Ordinary Least Squares (OLS), Weighted Least
Squares (WLS), Seemingly Unrelated Regression
(SUR), Two-Stage Least Squares (2SLS), Weighted
Two-Stage Least Squares (W2SLS) or Three-Stage
Least Squares (3SLS).
The wrapper fullfills the naming conventions of Rmetrics, returns a
S4 object named fSYSTEM
, and allows for timeSeries
objects
as input. In addition a S-Plus like FinMetrics function SUR
is made available.
The Models Based on 'systemfit' Include:
"OLS" | Ordinary Least Square Modelling, |
"WLS" | Weighted Least Square Modelling, |
"SUR" | Seemingly Unrelated Regression, |
"2SLS" | Two-Stage Least Squares, |
"W2SLS" | Weighted Two-Stage Least Squares, |
"3SLS" | Three-Stage Least Squares. |
Further Functions and Methods are:
print | S3 Print method for objects of class 'fSYSTEM', |
summary | S3 Summary method for objects of class 'fSYSTEM', |
predict | S3 Predict method for objects of class 'fSYSTEM'. |
S-Plus like Finmetrics Function:
SUR | A S-PLUS like function for "SUR" models. |
Note, that the contributed R package systemfit
is required!
If the package systemfit
is not installed on your computer
or not availalble for your operating system, then you can load it
as builtin function calling the internal Rmetrics function
systemfitBuiltin()
.
systemFit(formula, data = list(), method = c("OLS", "WLS", "SUR", "2SLS", "W2SLS", "3SLS", "W3SLS"), title = NULL, description = NULL, ...) nlsystemFit(formula, data = list(), method = c("OLS", "SUR", "2SLS", "3SLS"), start = NULL, title = NULL, description = NULL, ...) ## S3 method for class 'fSYSTEM': predict(object, newdata = object@data, se.fit = FALSE, se.pred = FALSE, interval = "none", ci = 0.95, ...) ## S3 method for class 'fSYSTEM': print(x, ...) ## S3 method for class 'fSYSTEM': summary(object, ...) ## S3 method for class 'fSYSTEM': coef(object, ...) ## S3 method for class 'fSYSTEM': fitted(object, ...) ## S3 method for class 'fSYSTEM': residuals(object, ...) SUR(formula, data = list(), ...)
ci |
[predict] - the confidence interval, by default 0.95. |
formula |
[systemFit] - the list of formulas describing the system of equations. |
data |
[systemFit] - the input data set in form of a data.frame or
timeSeries object.
|
description |
[systemFit] - a character string which allows for a brief description. |
interval |
[predict] - Type of interval calculation, one of "none" ,
"confidence" , or "prediction" .
|
method |
[systemFit] - a character string describing the desired method, one of: "OLS" , "WLS" , "SUR" , "2SLS" ,
"W2SLS" , "3SLS" , or "W3SLS" .
|
newdata |
[predict] - a new input data set in form of a data.frame or
timeSeries to be predicted.
|
object |
[predict][summary] - [coef][fitted][residuals][vcov] - an object of class fSYSTEM .
|
se.fit |
[predict] - a logical, should the standard error of the fitted values be returned? |
se.pred |
[predict] - a logical, should the standard error of the prediction be returned? |
start |
start values. |
title |
[systemFit] - a character string which allows for a project title. |
x |
[plot][print] - an object of class fSYSTEM .
|
... |
[systemFit] - additional optional arguments to be passed to the underlying function systemfit for the "OLS", "WLS", "SUR", "2SLS",
"W2SLS", "3SLS", or "W3SLS" method. These include: eqnlabels - an optional list of character vectors of names for the equation labels. formula3sls - formula for calculating the 3SLS estimator, one of "GLS" , "IV" , "GMM" , "Schmidt"
or "EViews" , see 'systemfit' details.
inst - one-sided model formula specifying instrumental variables or a list of one-sided model formulas if different instruments should be used for the different equations, only needed for "2SLS" , "W2SLS" and "3SLS" estimations.
maxiter - maximum number of iterations for "WLS" , "SUR" ,
"W2SLS" and "3SLS" estimations.
probdfsys - use the degrees of freedom of the whole system (in place of the degrees of freedom of the single equation) to calculate prob values for the t-test of individual parameters. q.restr - an optional j x 1 matrix to impose linear restrictions,
see R.restr ; default is a j x 1 matrix that
contains only zeros.
R.restr - an optional j x k matrix to impose linear restrictions
on the parameters by R.restr * b = q.restr ,
j = number of restrictions, k = number of all
parameters, b = vector of all parameters.
rcovformula - formula to calculate the estimated residual covariance matrix, see 'systemfit' details. saveMemory - save memory by omitting some calculation that are not crucial for the basic estimation, e.g. McElroy's R^2. single.eq.sigma - use different σ^2s for each single equation to calculate the covariance matrix and the standard errors of the coefficients, only "OLS" and "2SLS" .
solvetol - tolerance level for detecting linear dependencies when inverting a matrix or calculating a determinant, see see solve and det .
tol - tolerance level indicating when to stop the iteration, only "WLS" , "SUR" , "W2SLS" and
"3SLS" estimations.
TX - an optional matrix to transform the regressor matrix and, hence, also the coefficient vector, see 'systemfit' details. |
Ordinary Least Squares (OLS) estimates are biased and inconsistent
when endogenous variables appear as regressors in other equations
in the system. Furthermore, one observes that the errors of a set
of related regression equations are often correlated. Then the
efficiency of the estimates can in many cases be improved including
the correlations into the parameter estimation procedure. The
function eqnaFit
provides several methods which can
produce consistent and asymptotically efficient estimates for
systems of regression equations.
The variables in a system of equations can be characterized by
four types. These include Endogenous Variables which are
the variables determined by the system, Exogenous Variables
which are independent variables that do not depend on any of the
endogenous variables in the system, Predetermined Variables
which include both the exogenous variables and lagged endogenous
variables, which are past values of endogenous variables determined
at previous time periods, and Instrumental Variables which
are are predetermined variables used in obtaining predicted values
for the current period endogenous variables by a first-stage regression.
The use of instrumental variables characterizes estimation methods
such as two-stage least squares and three-stage least squares.
Instrumental variables estimation methods substitute these first-stage
predicted values for endogenous variables when they appear as
regressors in model equations.
Technical Details: 'systemfit'
The matrix TX
transforms the regressor matrix (X) by
X^{*} = X * TX
. Thus, the vector of coefficients is now
b = TX
cdot b^{*}, where b is the original
(stacked) vector of all coefficients and b^{*} is the new
coefficient vector that is estimated instead. Thus, the elements of
vector b are b_i = sum_j TX_{ij} cdot b^{*}_j.
The TX
matrix can be used to change the order of the
coefficients and also to restrict coefficients (if TX
has
less columns than it has rows). However restricting coefficients
by the TX
matrix is less powerfull and flexible than the
restriction by providing the R.restr
matrix and the
q.restr
vector. The advantage of restricting the coefficients
by the TX
matrix is that the matrix that is inverted for
estimation gets smaller by this procedure, while it gets larger
if the restrictions are imposed by R.restr
and q.restr
.
If iterated (WLS, SUR, W2SLS or 3SLS estimation with maxit
>1),
the convergence criterion is
sqrt{
sum_i (b_{i,g} - b_{i,g-1})^2 <=ft/
sum_i b_{i,g-1}^2 right. }
< tol
.
Here, b_{i,g} is the ith coefficient of the g-th iteration step.
The formula to calculate the estimated covariance matrix of the residuals, hat{Σ}, can be one of the following, see Judge et al., 1985, p. 469:
if rcovformula=0:
hat{σ}_{ij} =
(hat{e}_i' hat{e}_j) / T;
if rcovformula=1:
hat{σ}_{ij} =
(hat{e}_i' hat{e}_j) / sqrt{(T - k_i)*(T - k_j)};
if rcovformula=2:
hat{σ}_{ij} =
(hat{e}_i' hat{e}_j) / (T - k_i - k_j +
tr[(X_i'X_i)^{-1}X_i'X_j(X_j'X_j)^{-1}X_j'X_i].
If k_i = k_j, formula 1 and 2 are equal and yield an unbiased estimator for the residual covariance matrix. If k_i neq k_j, only formula 2 yields an unbiased estimator for the residual covariance matrix, but it is not neccessarily positive semidefinit and its inverse is not an unbiased estimator for the inverse of the residual covariance matrix. Thus, it is doubtful whether formula 2 is really superior to formula 1, see Theil, 1971, p. 322.
The formulas to calculate the 3SLS estimator lead to identical
results if the same instruments are used in all equations. If
different instruments are used in the different equations, only
the GMM-3SLS estimator, "GMM"
and the 3SLS estimator proposed
by Schmidt (1990), "Schmidt"
are consistent, whereas
"GMM"
is efficient relative to "Schmidt"
, see Schmidt,
1990.
Prediction:
The variance of the fitted values, used to calculate the standard errors of the fitted values and the confidence interval, is calculated by
Var[E[y^0]-hat{y}^0]=x^0 ; Var[b] ; {x^0}'
an the variances of the predicted values, used to calculate the standard errors of the predicted values and the prediction intervals, is calculated by
Var[y^0-hat{y}^0]=hat{σ}^2+x^0 ; Var[b] ; {x^0}'
Fit: Parameter Estimation
The function systemFit
returns an object of class "fSYSTEM"
with the following slots:
@call |
the matched function call. |
@data |
the input data in form of a data.frame or a
timeSeries object.
|
@description |
a character string which allows for a brief project description. |
@fit |
a summary of the results as a list returned from the underlying
functions from the systemfit package.
|
@formulas |
the list of formulas describing the system of equations. |
@method |
a character string describing the desired method, one of:
"OLS" , "WLS" , "SUR" , "2SLS" ,
"W2SLS" , "3SLS" , or "W3SLS" .
|
@title |
a character string which allows for a project title. |
coef |
the coefficients from an object of class fSYSTEM . A
one-column data frame of all estimated coefficients.
|
confint |
the confidence intervals of the coefficients of one equation
from an object of class fSYSTEM .
|
fitted |
the fitted values of all equations from an object of class
fSYSTEM .
|
residuals |
the residuals from an object of class fSYSTEM .
|
vcov |
the variance covariance matrix of all coefficients from an
object of class fSYSTEM .
|
method |
estimation method. |
g |
number of equations. |
n |
total number of observations. |
k |
total number of coefficients. |
ki |
total number of linear independent coefficients. |
df |
degrees of freedom of the whole system. |
iter |
number of iteration steps. |
b |
vector of all estimated coefficients. |
bt |
coefficient vector transformed by TX .
|
se |
estimated standard errors of b .
|
t |
t values for b .
|
p |
p values for b .
|
bcov |
estimated covariance matrix of b .
|
btcov |
covariance matrix of bt .
|
rcov |
estimated residual covariance matrix. |
drcov |
determinant of rcov .
|
rcovest |
residual covariance matrix used for estimation, only "SUR" and "3SLS". |
rcor |
estimated residual correlation matrix. |
olsr2 |
System OLS R-squared value. |
mcelr2 |
McElroys R-squared value for the system, only "SUR" and "3SLS". |
y |
vector of all (stacked) endogenous variables. |
x |
matrix of all (diagonally stacked) regressors. |
h |
matrix of all (diagonally stacked) instrumental variables, only "2SLS" and "3SLS". |
data |
data frame of the whole system including instruments. |
R.restr |
the restriction matrix. |
q.restr |
the restriction vector. |
TX |
matrix used to transform the regressor matrix. |
maxiter |
maximum number of iterations. |
tol |
tolerance level indicating when to stop the iteration. |
rcovformula |
formula to calculate the estimated residual covariance matrix. |
formula3sls |
formula for calculating the "3SLS" estimator. |
probdfsys |
system degrees of freedom to calculate prob values? |
single.eq.sigma |
different σ^2s for each single equation? |
solvetol |
tolerance level when inverting a matrix or calculating a determinant. |
eq |
a list that contains the results that belong to the individual equations. |
eqnlabel* |
the equation label of the i-th equation (from the labels list). |
formula* |
model formula of the i-th equation. |
inst* |
instruments of the i-th equation, only 2SLS and 3SLS. |
n* |
number of observations of the i-th equation. |
k* |
number of coefficients/regressors in the i-th equation (including the constant). |
ki* |
number of linear independent coefficients in the i-th equation
(including the constant differs from k only if there are
restrictions that are not cross-equation).
|
df* |
degrees of freedom of the i-th equation. |
b* |
estimated coefficients of the i-th equation. |
se* |
estimated standard errors of b .
|
t* |
t values for b .
|
p* |
p values for b .
|
covb* |
estimated covariance matrix of b .
|
y* |
vector of endogenous variable (response values) of the i-th equation. |
x* |
matrix of regressors (model matrix) of the i-th equation. |
h* |
matrix of instrumental variables of the i-th equation, only "2SLS" and "3SLS". |
data* |
data frame (including instruments) of the i-th equation. |
fitted* |
vector of fitted values of the i-th equation. |
residuals* |
vector of residuals of the i-th equation. |
ssr* |
sum of squared residuals of the i-th equation. |
mse* |
estimated variance of the residuals (mean of squared errors) of the i-th equation. |
s2* |
estimated variance of the residuals (hat{σ}^2) of the i-th equation. |
rmse* |
estimated standard error of the residulas (square root of mse) of the i-th equation. |
s* |
estimated standard error of the residuals (hat{σ}) of the i-th equation. |
r2* |
R-squared (coefficient of determination). |
adjr2* |
adjusted R-squared value. |
S3 Methods:
The output from the S3 summary
method prints the results
in form of a detailed report together with optional plots.
The output from the S3 print
method prints on object
of class fSYSTEM
.
The output from the S3 plot
method returns some diagnostic
plots.
S-Plus like SUR Function:
The function SUR
returns an object of class "fSYSTEM"
with the same slots returned by the function systemFit
for
method "SUR"
.
It is worth to remark, that there are two more R-packages which are of interest in this context:
The contributed R package "sem"
offers functions for fitting general
structural equation models by the method of maximum likelihood (SEM)
and for fitting a model by two-stage least squares (TSLS). This package
was written by John Fox.
The contributed R package "pls.pcr"
offers also functions for
multivariate regression. Principal Component Regression (PCR) and two
types of Partial Least Square Regression (PLS), simple-PLS and
kernel-PLS, are implemented by Ron Wehrens.
These two packages are not discussed here and are available from the CRAN server.
Wrapper for the nonlinear case are not yet available.
Jeff D. Hamann and Arne Henningsen for the systemfit
package,
Diethelm Wuertz for the Rmetrics R-port.
Greene W.H., (1993); Econometric Analysis, Second Edition, Macmillan.
Greene W.H., (2002); Econometric Analysis Fifth Edition, Prentice Hall.
Judge G.G., Griffiths W.E., Hill R.C, Lütkepohl H., Lee T.C., (1985); The Theory and Practice of Econometrics, Second Edition, Wiley.
Kmenta J., (1997); Elements of Econometrics, Second Edition, University of Michigan Publishing.
Schmidt P., (1990); Three-Stage Least Squares with different Instruments for different equations, Journal of Econometrics 43, p. 389–394.
Theil H., (1971); Principles of Econometrics, Wiley, New York.
lm
,
regFit
.
## SOURCE("fMultivar.3A-EquationsModelling") ## Not run: ## Note, "systemfit" is required: SYSTEMFIT = require(systemfit) if (SYSTEMFIT) { ## Examples from the 'systemfit' Package: data(kmenta) ## OLS Estimations: formulas = list(demand = q ~ p + d, supply = q ~ p + f + a ) FITOLS = systemFit(formulas, data = kmenta) FITOLS ## OLS Estimation with 2 Restrictions: Rrestr <- matrix(0, 2, 7) qrestr <- matrix(0, 2, 1) Rrestr[1,3] = 1 Rrestr[1,7] = -1 Rrestr[2,2] = -1 Rrestr[2,5] = 1 qrestr[2,1] = 0.5 FITOLS2 = systemFit(formulas, data = kmenta, R.restr = Rrestr, q.restr = qrestr) FITOLS2 ## Iterated SUR Estimation: FITSUR = systemFit(formulas, data = kmenta, method = "SUR", maxit = 100) FITSUR # Coefficients, Fitted Values, Residuals and Variance-Covariance Matrix: # Call by Method: coef(FITSUR) fitted(FITSUR) residuals(FITSUR) ## 2SLS Estimation: inst = ~ d + f + a FIT2SLS = systemFit(formulas, data = kmenta, method = "2SLS", inst = inst) FIT2SLS # Coefficients, Fitted Values, Residuals and Variance-Covariance Matrix: # Call by Slot: FIT2SLS@fit$coef FIT2SLS@fitted.values FIT2SLS@residuals ## 2SLS Estimation with Different Instruments in Each Equation: insts = list( ~ d + f, ~ d + f + a) FIT2SLS2 = systemFit(formulas, data = kmenta, method = "2SLS", inst = insts) FIT2SLS2 ## 3SLS Estimation with GMM-3SLS Formula: instruments = ~ d + f + a FIT3SLS = systemFit(formulas, data = kmenta, method = "3SLS", inst = instruments, formula3sls = "GMM") FIT3SLS } # if (SYSTEMFIT) ## SEE ALSO: # Demo File: xmpEqnsGrunfeld.R # Estimation of Grunfeld's Model Data with OLS and SUR ## End(Not run)