bng_panic#

Purpose#

Computes the Pe test on ADF p-values found in Bai & Ng (2004) using panel analysis of idiosyncratic and common components (PANIC) test of nonstationarity.

Format#

{ ADFe, pval, lags, Pe, n_factors } = bng_panic(y, model[, pmax, ic_lags, kmax, ic_factors])#
Parameters:

  • y (TxN matrix) – Panel data to be tested.

  • model (Scalar) –

    Model to be implemented.

    1

    Constant.

    2

    Constant and trend.

  • pmax (Scalar) – Optional, the maximum number of lags for \(\Delta y\). Default = 8.

  • ic_lags (Scalar) –

    Optional, the information criterion used for choosing lags. Default = 3.

    1

    Akaike.

    2

    Schwarz.

    3

    t-stat significance.

  • kmax (Scalar) – Maximum number of factors. Default = 5.

  • ic_factors (Scalar) –

    Information Criterion for optimal number of factors. Default = 1.

    1

    PCp criterion.

    2

    ICp criterion.

Returns:

  • ADFe (Scalar) – ADF statistic for idiosyncratic components for each cross-section.

  • pval (Scalar) – p-value of ADFe.

  • lags (Scalar) – Number of lags selected by chosen information criterion.

  • Pe (Scalar) – Pe statistic based on principal components with N(0,1).

  • n_factors (Scalar) – Number of factors by chosen information criterion

Examples#

library tspdlib;

// Load date file
y = loadd(__FILE_DIR $+ "PDe.dat");

/*
** Using the defaults
** for maximum number of lags,
** information criterions,
** and maximum number of factors.
*/

/*
** Model with constant
*/
model = 1;
{ ADFe, pval, lags, Pe, nf } = bng_panic(y, model);

Source#

pd_panic.src

See also

Functions bng_panicnew(), jwl_panicadj(), jwr_panicca(), pd_stationary()

On this page