pd_kpss

Purpose

Computes the panel data KPSS test.

Format

{ testd_hom, testd_het, m_lee_est, brks } = pd_kpss(y, model[, nbreak, bwl, varm, pmax, bCtl])

Parameters:

• y (TxN matrix) – Wide format panel data.

• model (Scalar) –

  Model to be implemented.

  1	Constant (Hadri test).
  2	Constant and trend (Hadri test).
  3	Shift in the mean.
  4	Shift in mean and trend.

• nbreak (Scalar) – Optional, number of breaks to consider (up to 5). Default = 5.

• bwl (Scalar) – Optional, bandwidth for the spectral window. 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

Returns:

• test_hom (Scalar) – Panel test statistic assuming homogenous variance.

• test_het (Scalar) – Panel test statistic assuming heterogenous variance.

• m_lee_est (Matrix) – Contains results for each individual group. First column contains KPSS test statistic, second column contains the optimal number of lags selected using the mlwz criteria.

• brks (Scalar) – Estimated breaks. Breaks for each individual group are contained in separate rows.

Examples

new;
cls;
library tspdlib;

// Load data
data = loadd(__FILE_DIR $+ "pd_gdef.gdat");
data = setColDateFormats(data, "%Y", "Year");

// Set model to  hve break in constant and trend
model = 4;

// Compute panel data test
{ test_hom, test_het, kpsstest, m_br} = pd_kpss(data, model);

// Plot results
plotPDKPSS(data, m_br, kpsstest);

Source

pd_kpss.src

See also

Functions kpss_1break(), kpss_2breaks(), lmkpss()