cusum
=====

Purpose
-------
Compute the Brown, Durbin, and Evans (1975) CUSUM test using the empirical fluctuation process of the cumulative sums of standardized residuals.



Format
------
.. function::  { cs, cst2 } = cusum(yt, xt [, gr, minwin]);


  :param yt: data.
  :type yt: vector

  :return xt: estimation regressors.
  :rtype xt: matrix

  :param gr: Optional, indicator to graph vector of CUSUM test statistics. 0 for no printing, otherwise graphs and printed. Default = 1.
  :type gr: scalar

  :param gr: Optional, minimum estimation window size. If :math:`0 \lt minwin \lt 1`, minwin indicates a percentage of the number of observations in yt. If :math:`minwin \gt 1`, minwin indicates the actual number of observations in the window size. Default = 0.3.
  :type gr: scalar

  :return cst: CUSUM test statistics.
  :rtype cst: vector

  :return cst2: CUSUM squared test statistics.
  :rtype cst2: vector

Example
-------
::

  new;
  cls;
  library tsmt;

  /********************************************/
 // This generates 300 observations of an
 // linear time series with a break in the constant
 // at observations 90

 b1 = { 0.6, 0.25, 0.75 };
 b2 = { 0.95, 0.25, 0.75 };

 n1 = 90;
 n_tot = 300;
 xt = ones(n_tot, 1)~rndn(n_tot, 2);
 et = rndn(n_tot, 1);

 // Create series with break
 y1 = xt[1:n1, .]*b1 + et[1:n1, .];
 y2 = xt[n1+1:n_tot, .]*b2 + et[n1+1:n_tot, .];
 yt_break = y1|y2;

 // Create series without break
 yt = xt*b1 + et;

 /********************************************/
 //Next set the cusum parameters
 /********************************************/
 minwin = 30;
 gr = 1;

 /********************************************/
 //Test residuals using cusum
 /********************************************/
 plotOpenWindow();
 { cst1, cst1_2 } = cusum(yt_break, xt);

 plotOpenWindow();
 { cst2, cst2_2 } = cusum(yt, xt);

Remarks
-------
The Brown, Durbin, and Evans (1975) CUSUM test considers the empirical fluctuation process of the cumulative sums of standardized residuals. Under the null hypothesis of constant coefficients the residuals should have zero mean. Hence, significant deviation from zero at time indicates possible structural change at time *t*. The test is valid for dynamic models.

Reference
---------
Brown, R.L., Durbin, J., and Evans, J.M. (1975). Techniques for testing the constancy of regression relationships over time, Journal of Royal Statistical Society, Series B, 35, 149-192. .

Library
-------
tsmt

Source
------
cusum.src