Two reduced-form approaches to the derivation of the maximum-likelihood estimators for simultaneous-equation systems