Estimation of the constrained parameters via the EM-type algorithm
Hong Liu · Journal of Northeast Normal University · 2008
This paper considers the problem of parameters estimate in multivariate normal models.By using the EM and ECM algorithm,the maximum likelihood estimate of the parameters which are restricted to the simple ordering,the umbrella ordering or the increasing convex ordering,are given whenever the covariance matrix is known or not.If the dimension of the parameters is less than or equal to three,the explicit estimates are proven.Otherwise,if the dimension is higher,the corresponding linear transform that is used to get the maximum likelihood estimate of the restricted parameters is shown.