Effect of Autocorrelated Observations on Confidence Sets Based upon Chi-Square Statistics

How the presence of autocorrelation in a multivariate normal sample affects the confidence level of confidence regions based upon chi-square statistics is investigated. Particular consideration is given to the effect that this violation of the assumption of a random sample has upon the construction of confidence regions for the mean ¿ and the scalar ¿2, in ¿2V, where V is a known dispersion matrix.