Estimation for the multivariate errors-in-variables model with estimated error covariance matrix (Q796217)

The authors consider the estimation problem of multivariate errors-in- variables models. Let \(r\times 1\) row vectors \(y_ i\) and \(k\times 1\) row vectors \(x_ i\) satisfy \(y_ i=\beta_ 0+x_ i\beta\), \(i=1,2,...,n\), where \(\beta_ 0\) and \(\beta\) are 1\(\times r\) and \(k\times r\) matrices of parameters, respectively. We observe \(Y_ i\) and \(X_ i\), which satisfy \(Y_ i=y_ i+e_ i\), \(X_ i=x_ i+u_ i\), \(i=1,2,...,n\). \(\epsilon_ i=(e_ i,u_ i)\) are assumed to be i.i.d. with mean zero and covariance matrix \(\Sigma_{\epsilon \epsilon}\). Assume that the \(\epsilon_ i\) are independent of the \(x_ j\) for all i and j and an estimator \(S_{\epsilon \epsilon}\) of \(\Sigma_{\epsilon \epsilon}\) is available. Under the assumption of joint normal distribution of \(x_ j\) and \(\epsilon_ i\), the authors derive the maximum likelihood estimators of \(\beta_ 0\), \(\beta\), \(\Sigma_{\epsilon \epsilon}\), \(Cov(X_ i)=\Sigma_{xx}\) and \(E(x_ i)\). Limiting properties and strong consistency of the estimators are obtained for a wide range of assumptions.