Covariance hypothesis which are linear in both the covariance and the inverse covariance (Q1107929)

The author has studied the structure of statistical hypotheses for the family of normal distributions, which are linear in both the covariance and the inverse covariance. It is shown that such hypotheses are products of models each of which consist of either i.i.d. random vectors which have a covariance with a real, complex or quaternion structure or i.i.d. random vectors with a parameterization of the covariance which is given by the Clifford algebra.