Abstract
A typical statistic encountered can be characterized as a ratio of polynomials of arbitrary degree in a random vector. This vector may possess any admissible cumulant structure. We provide in this paper general formulae for the effect of nonnormality on the density and distribution functions of this ratio. The results appear in terms of generalized cumulants, a theory developed by McCullagh (1984, Biometrika 71, 461-476). With the aid of suitable notation, the expressions are applied to the distributions of tests for heteroskedasticity and autocorrelation, the least-squares estimator of the autoregressive coefficient in a dynamic model, and tests for linear restrictions.
| Original language | English |
|---|---|
| Pages (from-to) | 52-78 |
| Number of pages | 27 |
| Journal | Econometric Theory |
| Volume | 13 |
| Issue number | 1 |
| DOIs | |
| State | Published - 1997 |
| Externally published | Yes |