This file contains instructions for installation and use of the procedures adecm.g and jmle.g for

use in estimating the error correction matrix and cointegrating matrix in an Error Correction

Model.
Authors: Douglas Hodgson and Keith Vorkink, University of Rochester
Installation:
Copy the two .g files into the /gauss/src directory.  They will be directly available by calling

into your program.  There is no need to edit/add libraries.

Purpose:

The procedure adecm.g is the main procedure to call to estimate an Error Correction Model.

Within this procedure jmle.g is called to obtain johansen's MLE preliminary estimators

which are used to obtain the adaptive estimators.  We note that jmle.g can be called independently

to estimate the cointegrating basis and error correction matrix.  The main reference for this

procedure is Douglas Hodgson, (95), "Adaptive Estimation of Error Correction Models" RCER working

paper #410.  The reader is also referred to James Hamilton's Time Series Analysis book for

Error Correction Representation and Johansen's MLE.

Format:

{b,vb} = adecm(y,k);

Inputs:

                y                nxr matrix, series where cointegration is assumed present

                x                nx(q-r) matrix, series assumed to have unit roots

Outputs:

b                vector or coefficients of A, PSI, B (see Hodgson for representation)

                vb                covariance matrix of parameter estimates

Remarks:
  The models assumes the distribution is symmetric and proceeds to estimate

                                the residual distribution nonparametrically.  This requres trimming and smoothing

                                parameters that can be set to default values or specified by the user.  Addititionally

                                the model can be estimated with or without a constant, this is chosen interactively

                                in the inputs procedure which also determines the lags to include in the model, the

                                trimming parameter, and the smoothing parameter.

Example:  
loadm dat[658,5] = c:\gauss\data\cdxr1.txt;
fxr = ln(dat[1:658,2]);
sxr = ln(dat[1:658,5]);
y = sxr;
x = fxr;          
{v,b} = adecm(y,x);