Power of Heteroskedasticity Tests in Presence of Various Types of Skedastic Function and Sample Size

摘要

This paper presents evidence from Monte Carlo simulation aimed at exploring power of heteroskedasticity tests to verify presence of various patterns of non-constant variances in linear models. Three independent MC simulations were set up to investigate impact of error term variance changing with power of the proportionality factor (schemes 1 and 2) or in response to variance multiplication constant applied to a subset of data (scheme 3). Empirical power was shown to vary in accordance with the simulated form of the skedastic function and sample size. Tests mainly showed adequate power for the skedastic function in scheme 1 and to a lesser extent for variance function in scheme 3. On the other hand, skedastic functions simulated in scheme 2 were the most difficult to detect by most heteroskedasticity tests.