EFFECT OF NON-NORMALITY ON INFERENCES ABOUT VARIANCE COMPONENTS.
WISCONSIN UNIV MADISON DEPT OF STATISTICS
Pagination or Media Count:
Bayesian methods are used to analyse the one-way random effect model y sub ij mu a sub i e sub ij in which e sub ij are assumed normal, N0, sigma squared sub e, and a sub i are assumed to have a mixture of two normals, .95N-.05phik-1 sigma, sigma squared .05N.95phik-1 sigma, k squared, sigma squared. It is shown that for moderately sized sample, inferences regarding sigma sigma squared sub e vare sub ij are insensitive but those of sigma squared sub a vara sub i are very sensitive to changes of the two non-normality parameters k, phi. Author
- Statistics and Probability