Właściwości wybranych metod małopróbkowego wnioskowania o rzędzie kointegracji
Small Sample Inference on Cointegration Rank

Summary/Abstract: In the paper, properties of small sample cointegration rank tests are compared. The Monte Carlo experiments conducted for different data generating processes and areas in the parametric space indicate that (i) the likelihood ratio test with Bartlett correction usually have the best properties, (ii) bootstrapping the Bartlett corrected test does not lead to improvement of test properties, (iii) size of asymptotic test and tests with degrees-of-freedom correction is usually heavily distorted, (iv) the Bartlett correction can lead to a small improvement of power, (v) correlation of error terms between stationary and non-stationary component of canonical form lead to a signifi cant increase of power and size of test, but only size of Bartlett corrected test and bootstrapped test converge to nominal size.

• Details
• Contents