coint_cissanso#

Purpose#

Computes a LM-type statistic to test the null hypothesis of cointegration allowing for the possibility of a structural break.

Format#

{ SCols, TBols, SCdols, TBdols, lambda, cv } = coint_cissanso(y, x, model[, bwl, varm, trimm, q])#
Parameters:
  • y (Nx1 matrix) – Dependent variable.

  • x (NxK matrix) – Independent variable.

  • model (Scalar) –

    Model to be implemented.

    1

    Model An: Level shift.

    2

    Model A: Level shift with trend.

    3

    Model D: Regime shift.

    4

    Model E: Regime and trend shift.

  • bwl (Scalar) – Optional, bandwidth length for long-run variance computation. Default = round(4 * (T/100)^(2/9)).

  • varm (Scalar) –

    Optional, long-run consistent variance estimation method. Default = 1.

    1

    iid.

    2

    Bartlett.

    3

    Quadratic Spectral (QS).

    4

    SPC with Bartlett (Sul, Phillips & Choi, 2005)

    5

    SPC with QS

    6

    Kurozumi with Bartlett

    7

    Kurozumi with QS

  • trimm (Scalar) – Optional, trimming rate. Default = 0.10.

  • q (Scalar) – Optional, number of leads and lags for DOLS estimation. Default = int(4 * (t/100)^(2/9)).

Returns:
  • SCols (Scalar) – SC test based on OLS estimation

  • TBols (Scalar) – Break location based on OLS estimation.

  • SDols (Scalar) – SC test based on DOLS estimation

  • TBDols (Scalar) – Break location based on DOLS estimation.

  • lamdba – Fraction of break (TB/T)

  • cv (Vector) – 1%, 5%, 10% critical values for the chosen model

Examples#

library tspdlib;

// Load dataset
data = loadd(getGAUSSHome() $+ "pkgs/tspdlib/examples/ts_coint.csv",
                                 ". + date($Date, '%b-%y')");

// Define y and x matrix
y = data[., 1];
x = data[., 2:cols(data)];

// Level shifts
model = 1;

// Call test
{ SCols, TBols, SCdols, TBdols, lambda, cv } = coint_cissanso(y, x, model);

Source#

coint_cissano.src