On structural comparative calibration under a \(t\)-model (Q1966006)

The authors consider the problem of assessing relative calibrations and relative accuracies of a set of \(p\) instruments each designed to measure the same characteristic on a common group of \(n\) individuals by using the \(t\)-distribution. Two models have been proposed to analyse data from such experiments: the regression model version and the factor analysis version. All these models assume errors normally distributed. In this paper the normal distribution is replaced by the \(t\)-distribution. Estimation is approached via the EM-algorithm. The method allows to pass estimators and the information matrix from one model to the other. Some numerical illustrations of the procedures are presented. To derive the information matrix, the authors present some results on expectations of certain functions of normal and \(t\)-random variables.